1

Survey on Multi-Access Edge Computing forInternet of Things Realization

Pawani Porambage Student Member IEEE Jude Okwuibe Student Member IEEEMadhusanka Liyanage Member IEEE Mika Ylianttila Senior Member IEEE and Tarik Taleb Senior

Member IEEE

AbstractmdashThe Internet of Things (IoT) has recently advancedfrom an experimental technology to what will become thebackbone of future customer value for both product and servicesector businesses This underscores the cardinal role of IoT on thejourney towards the fifth generation (5G) of wireless communi-cation systems IoT technologies augmented with intelligent andbig data analytics are expected to rapidly change the landscapeof myriads of application domains ranging from health care tosmart cities and industrial automations The emergence of Multi-Access Edge Computing (MEC) technology aims at extendingcloud computing capabilities to the edge of the radio accessnetwork hence providing real-time high-bandwidth low-latencyaccess to radio network resources IoT is identified as a key usecase of MEC given MECrsquos ability to provide cloud platformand gateway services at the network edge MEC will inspire thedevelopment of myriads of applications and services with demandfor ultra low latency and high Quality of Service (QoS) due to itsdense geographical distribution and wide support for mobilityMEC is therefore an important enabler of IoT applications andservices which require real-time operations In this survey weprovide a holistic overview on the exploitation of MEC technologyfor the realization of IoT applications and their synergies Wefurther discuss the technical aspects of enabling MEC in IoT andprovide some insight into various other integration technologiestherein

Index TermsmdashMulti-Access Edge Computing (MEC) Internetof Things (IoT) 5G edge computing virtualization networkarchitecture latency reliability

I INTRODUCTION

OVER the last four decades the Internet has evolvedfrom peer-to-peer networking to world-wide-web and

mobile-Internet to the Internet of Things (IoT) (Figure 1) IoTemerged as a huge paradigm shift by connecting a versatile andmassive collection of smart objects to the Internet With IoTpeople and things are able to connect at any time to any placewith anything and anyone ideally using any path or networkand any available services [1] From the user and applicationpoints of view fifth generation (5G) wireless networks willbe highly capable mobile networks with high bandwidth (eg10 Gbps) very low latency (eg 1 ms) and low operationalcost which will lead to highly improved quality of serviceand quality of experience Another significant advancement ofthe Internet will be the Tactile Internet which is a highly

Pawani Porambage Jude Okwuibe Madhusanka Liyanage and MikaYlianttila are with the Center for Wireless Communications University ofOulu Finland e-mailfirstnamelastnameoulufi

Tarik Taleb is with Department of Communications and Networking AaltoUniversity Finland and Sejong University Korea e-mail tariktalebaalto

advanced use case of human-to-machine and machine-to-machine interaction characterized by ultra low latency withextremely high availability reliability and security

Fig 1 Evolution of the Internet

IoT system is poised to induce a significant surge in demandfor data computing resources as well as networking infras-tructures in order to accommodate the anticipated myriads ofinterconnected devices Meeting these extreme demands willnecessitate a modification to existing network infrastructuresas well as cloud computing technologies

Mobile Edge Computing was introduced by the EuropeanTelecommunications Standards Institute (ETSI) Industry Spec-ification Group (ISG) as a means of extending intelligence tothe edge of the network along with higher processing andstorage capabilities [2] From 2017 the ETSI industry grouprenamed it to Multi-Access Edge Computing (MEC) since thebenefits of MEC technology reached beyond mobile and intoWi-Fi and fixed access technologies Nevertheless the namechange conveniently allows ETSI to retain the MEC acronymwhich has become widely recognized among stakeholders inthe industry

The underlying principle of MEC is to extend cloud comput-ing capabilities to the edge of cellular networks This will min-imize network congestion and improve resource optimizationuser experience and the overall performance of the networkBy leveraging on the Radio Access Networks (RANs) MECwill improve heavily on latency and bandwidth utilizationmaking it easier for both application developers and contentproviders to access network services Several technologies areidentified as enabling technologies for MEC realization theseinclude Software Defined Networking (SDN) Network Func-tion Virtualization (NFV) Information Centric Networking(ICN) and Network Slicing

arX

iv1

805

0669

5v1

[cs

NI]

17

May

201

8

2

A Role of MEC for IoT

Generally cloud computing enables the outsourcing ofstorage and processing functionalities of IoT data to a thirdparty in order to ease the hazel involved in self-managementand data protection However the centralized nature of con-ventional cloud servers may face several challenges such as thesingle point of failure lack of location awareness reachabilityand latencies associated with typical Wide Area Networks(WANs) On the other hand many IoT applications need tobe served with decentralized systems which need mobilitymanagement geo-distribution location awareness scalabilityand ultra-low latency Mission critical communication IoTuse cases need latency as low as 1 ms and reliability ashigh as 9999 For instance factory automation applicationsmay typically require a reliability of 10minus9 packet loss rateand a latency range of 250 micros to 10 ms [3] Thereforethe conjugation of IoT applications and centralized cloudservers may introduce several limitations and vulnerabilitiesIn addition the rapid growth of IoT devices and big data setsmay also create cumbersome traffic on telecommunicationsnetworks

Edge computing was conceived in a bid to fill the gapbetween the centralized cloud and IoT devices Apart fromMEC there are other edge computing paradigms such as Mo-bile Cloud Computing (MCC) fog computing and cloudletsThey tend to coexist with MEC in many technical contextshence the tendency for a misappropriation of these technolo-gies given that they all have similar origin However thesetechnologies are intrinsically different and each of them comeswith its unique value proposition to both existing and futuremobile networks as summarized in Table I

ETSI has identified IoT as one of the key use cases ofMEC [2] MEC has opened many new frontiers for networkoperators service and content providers to deploy versatileand uninterrupted services on IoT applications MEC and IoTfacilitate each other with mutual advantages MEC empowerstiny IoT devices with significant additional computationalcapabilities through computation offloading Similarly IoTexpands MEC services to all types of smart objects rangingfrom sensors and actuators to smart vehicles As shown inFigure 2 MEC servers can perform as gateway nodes whichcan aggregate and process the small data packets generatedby IoT services before they reach the core network Assummarized in [4] the three key benefits of the collaboration

between IoT and MEC are 1) lowering the amount of trafficpassing through the infrastructure 2) reducing the latencyfor applications and services and 3) scaling network servicesdiversely Among these the most significant is the low latencyintroduced by MEC due the reduced physical and virtualcommunication distance

Fig 2 IoT gateway service scenario [2]

B Paper motivation

At present IoT has become a fairly mature technology Asa result the recent decade has seen a plethora of surveyspublished in multiple research areas on IoT including enablingconcepts [5] visions and challenges [6] technologies [7]standardization [8] architecture [9] security [10] [11] pri-vacy [12] trust [13] Social Internet of Things (SIoT) [14]communication [15] context awareness [16] and future direc-tions [6] [17] Few other papers are focused on the combinedaspects of IoT research and their potential application sce-narios [7] [18]ndash[20] Some of these surveys were publishedduring the time when IoT was more of a visionary paradigmthan a real world platform Many future research possibilitiesdiscussed in those papers have already been achieved andcommercialized with high market values However there is yetto be a sufficient number of publications on MEC technologygiven that is relatively a novel technology which lies at theintersection of mobile cloud computing and wireless commu-nication In Table II we summarize the recently published sur-veys on MEC These articles are focused on MEC taxonomyfuture research directions and more specific MEC attributessuch as communication computation offloading security andvirtualization These studies are quite shallow in addressingthe MEC integration with IoT they are mostly focusing onthe requirements and usability of MEC in IoT applicationsIn this short magazine article [21] the authors discuss the

TABLE I High level comparison of edge computing paradigms

MEC Fog computing Cloudlet MCCInitial promotion ETSI (2014) Cisco (2011) Carnegie Mellon Uni (2013) Aepona (2010)Objective Bring cloud computing capabilities closer to User Equipment (UE)Infrastructure owners Telecom operator Private entities individualsNode location Radio network controller or

macro base stationAny strategic location between end user device and cloud

SW architecture Mobile orchestrator based Fog abstraction layer based Cloudlet agent based Service orientedService accessibility Direct access from the closest UE Via Internet connectionLatency and jitter Low HighContext awareness High Medium Low HighStorage capacity andcomputation power

Limited High

Relevance to IoT High Low

3

examples of MEC deployment with special reference to IoTuse cases

To the best of our knowledge there is not a single surveywhich addresses broader range of areas about MEC and itsinfluence on IoT realization Since both MEC and IoT arevery essential to the realization of 5G it is vital to expresstheir associativity in terms of application scenarios and keytechnical attributes Our goal is to broaden the horizons ofpotential inter-dependencies of MEC and IoT technologies andtheir related applications in future 5G and beyond

Furthermore in our previous survey [4] we discuss the roleof MEC in 5G network edge cloud architecture and orchestra-tion There we do not explicitly address the integration of MECfor the realization of IoT and related applications In additionto MEC integration technologies like SDN NFV and networkslicing discussed in [4] we consider ICN in this work There-fore this survey sets to provide a comprehensive overviewof the state-of-the-art technologies which are required for thecomplementary integration of MEC with IoT In this surveyour contributions manifold into three main categories

1) Providing a comprehensive survey on the exploitationof MEC technology for the realization of different IoTapplications

2) Presenting a holistic overview of related works andthe future research directions in areas of scalabilitycommunication computation offloading resource allo-cation mobility management security privacy and trustmanagement of MEC-IoT integration

3) Providing a concise summary of the state-of-the-artMEC integrating technologies for IoT and relatedprojects

C Paper organization

The rest of the paper is organized as follows Section II sum-marizes the well-known IoT applications that require a note-worthy assistance of MEC like edge computing technologiesSection III is particularly focused on technological aspects ofMEC enabled IoT systems in terms of scalability communica-tion computation offloading resource management mobilitymanagement security privacy and trust management Eachtechnical aspect is described with its requirements and relatedworks Section IV and V respectively summarize the relatedwork on different MEC integration technologies and the pro-ceeding research projects in the respective areas Section VIdescribes the lessons learned and the future research directions

TABLE II Summary of important surveys on MEC

Aspect Ref Main contribution Relevance to IoT

Research directions

[22] An elaboration of edge-centric vision and its future re-search challenges

No explicit focus on IoT

[23] A comprehensive overview on sate-of-the-art and futureresearch directions for MEC

Concisely describes how MEC can improve latencyand support big data handling in different IoT deploy-ments

[24] A presentation of MEC related definitions applicationsopportunities and research challenges

Provides no detailed description on IoT Identifies IoTdata handling as a key use case of MEC

[25] A concise tutorial of three edge computing technologiesincluding MEC cloudlets and fog computing

Describes the exploitation of edge computing tech-nologies for IoT with respect to standardization effortsprinciples architectures and applications

[26] A comprehensive survey of relevant research and techno-logical developments in the area of MEC

Identifies MEC services for IoT big-data analytics

Taxonomy [27] A taxonomy of MEC based on different aspects includingits characteristics access technologies applications andobjectives

Classifies MEC applications as computational offload-ing collaborative computing memory replication inIoT and content delivery

[28] A classification of applications deployed in MEC systems No explicit focus on IoTArchitecture andComputationOffloading

[29] A detailed study on decision on computation offloadingallocation of computing resources and mobility man-agement along with a summary of MEC use cases andstandardization efforts

Describes MEC acting as an IoT gateway

Virtualization[4] A survey of 5G network edge cloud architecture and

orchestration with a summary of MEC virtualization tech-nologies including Virtual Machines (VMs) SDN NFVand network slicing

Explains how MEC platform can encompass a localIoT gateway functionality capable of performing dataaggregation and big data analytics for applicationdomains

[30] An investigation on how to exploit SDN for enabling edgecomputing

Discuses SDN scenarios based on IoT and edge Com-puting and the future research

[31] An elaboration of network slicing from an E2E perspectiveon principles enabling technologies and solutions

Describes the role of massive IoT as a key use caseof 5G and network slicing

Communication [32] An comprehensive survey on joint radio-and-computational resource management in MEC systems

Briefly introduces the role of MEC in IoT

[33] A comprehensive survey of issues on computing cachingand communication techniques in MEC

Describes specific applications and use cases of MECin IoT including healthcare wireless sensor systemssmart grid smart home and smart city

MEC-IoT [21] An overview about the role of MEC in IoT use cases Provides examples of MEC deployments for IoT casesSecurity safety and data analytics Vehicle to infras-tructure communication Computation offloading toedge cloud

Security [34] A discussion of the security threats and challenges in theedge paradigms along with the promising solution foreach specific challenge

No explicit discussion on IoT Briefly discusses howIoT will benefit from edge computing and relatedsecurity threats

4

TABLE III Summary of important acronyms

Acronym Definition Acronym Definition

3GPP Third Generation Partnership Project 5G Fifth Generation Wireless NetworkAI Artifical Intelligence AR Augmented RealityBLE Bluetooth Low Energy CaPC Cloud-aware Power ControlCPS Cyber Physical System C-RAN Cloud Radio Access NetworkD2D Device-to-device DDoS Distributed Denial of ServiceDoS Denial of Service E2E End-to-endEC Edge Computing eMBB enhance Mobile BroadbandEMM Energy-aware Mobility Management eNodeB Evolved Node BETSI European Telecommunications Standards Institute EU European UnionFiWi Fiber-enable Wireless F-RAN Fog Radio Access NetworkGDPR General Data Protection Regulation ICN Information Centric NetworkingICT Information Communication Technology IIoT Industrial Internet of ThingsIoT Internet of Things ISG Industry Specification GroupKDN Knowledge-Defined Networking LPWAN Low-power Wide Area NetworkLTE Long Term Evolution M2M Machine-to-machineMANO Management and Orchestration MCC Mobile Cloud ComputingMEC Multi-Access Edge Computing MIFaaS Mobile-IoT-Federation-as-a-ServiceMitM Man-in-the-Middle mmW millimeter-WaveMR Mixed Reality NB-IoT Narrow-band IoTNFV Network Function Virtualization PbD Privacy by DesignQoE Quality of Experience QoS Quality of ServiceRAN Radio Access Networks RAT Radio Access TechnologyRFID Radio-Frequency Identification RNC Radio Network ControllerSCeNB Small Cell eNodeBs SDLB Software Load BalancerSDN Software Defined Networking SDP Software Defined PrivacySIoT Social Internet of Things TDMA Time-division Multiple AccessUAV Unmanned Aerial Vehicles UE User EquipmentV2V Vehicle to Vehicle V2X Vehicle to EverythingVANET Vehicular Ad-hoc Network VM Virtual MachineVNF Virtual Network Function VR Virtual RealityVRARA Virtual RealityAugmented Reality Association WAN Wide Area NetworkingWAP Wireless Access Point WIoT Wearable Internet of ThingsWLAN Wirless Local Area Networking WSN Wireless Sensor Network

Finally Section VII concludes the paper We provide thedefinitions of frequently used acronyms in Table III

II IOT AND MEC APPLICATION SCENARIOS

This section focuses on how IoT can leverage MEC tech-nology in various application scenarios IoT itself is a classicapplication of MEC where the key value proposition of MECis exemplified in a variety of application scenarios (Figure 3)These values become evident in the utility factor measured bythe end user experience while using such IoT related services

Table IV and V respectively show the characteristics of dif-ferent IoT applications and how each application benefits fromMEC-IoT integration In addition Table VI summarizes thereviewed state-of-the-art applications in MEC-IoT domains

A Smart home and Smart city

One of the pioneering applications of the IoT technologyhas been in the areas of home automation and consumerelectronics [39] Several smart home applications that are builton the basis of IoT concept are already available in mostconsumer markets These range from the simple thermostatsensors to other more sophisticated automation systems likesmart metering smart heating and lighting cleaning servicesand home entertainment systems That notwithstanding theamount of data that would be generated on a typical IoTnetwork like the smart home is expected to be huge Hencetransferring such data to the centralized cloud servers will be

impractical with most pre-MEC techniques As a solutionMEC leverages specialized and reliable local services forprocessing and storage capabilities for the large IoT trafficcreated within a building The conventional gateways whichallow IoT applications to run on the centralized cloud can beempowered with MEC-server functionalities [40] [41] Thisextends gateway functionalities to the edge of the networkwith reduced communication latency Since such appliancesare statically deployed in smart home or smart building envi-ronments the cooperation with MEC servers will offer someother features such as easy instantiation relocation privacypreservation and upgrading when necessary [21] [42]

Correspondingly IoT technology has advanced from hometo community and even city scale applications We see nu-merous future promises for public safety health care utilitytourism and the transport sectors Enormous IoT data trafficproduced in smart cities can be ideally processed at the edgeof the network providing low latency and location aware-ness [43] [44] In particular a video cameras (ie deployedfor surveillance) connected with a Long Term Evolution (LTE)network can convey video streams to the MEC server forreal-time processing and anomaly detection [21] Collaborativeedge paradigms that connect multiple MEC servers (iededicated for different services) will advocate the applicationswhich need to process geographically distributed data Forinstance a connected health care application requires to col-laborate with entities from multiple domains such as hospitalpharmacy insurance logistics and government [45]

5

Fig 3 IoT and MEC application scenarios

TABLE IV Characteristics of Different IoT application

IoT Application Data type Data Capacity BackhaulConnectivity

Expected latency Number of IoT Devices

Smart home Stream Historical data

ge 10 MB of data per house-hold per day

Realtime 1 ms -1000 s ge10-100 per house

Smart city Stream Massive data

ge10-100 million GB of dataper city per day

Realtime le1ms ge1000-1million per city

Remote surgery [35] Stream data ge15 million per year Realtime le200 ms ge10-100 per surgeryRemote consultancy Stream data ge 500 million visits per year Realtime 1 ms-100 s 1-10 per appointmentAutonomous vehicles Stream

Massive datage 100 GB per vehicle per day Realtime le1 ms 50-200 per vehicle

AR [36] Stream Massive data

ge1 GBps Realtime le1 ms ge02 million globally

VR [36] Stream Massive data

ge1 GBps Realtime le1ms ge02 million globally

Gaming [36] Stream Massive data

ge10 Mbps Realtime le10 ms ge1 billion globally

Retail [37] Stream Historical data

100 Mbps - 1 Gbps RealtimeIntermittent

le1 ms ge100-1000 per shop

WIoT Stream data lt 1 GB per device Intermittent Several Hours ge1-10 per personFarming Historical data ge 1 GB per farm Intermittent Several hours 100-100000 per farmSmart energy Stream

Massive datage 100000 GB per day Realtime

Intermittent1ms - 10 mins ge 1 billion per grid

Industrial Internet [38] Stream Massive data

ge 100000 GB per day Realtime le1 ms ge 1 million per factory

6

TABLE V MEC and IoT benefits for each application

Required characteristicsof MEC and IoT

Description Smar

tho

me

Smar

tci

ty

Rem

ote

surg

ery

Rem

ote

heal

thco

nsul

tanc

y

Aut

onom

ous

vehi

cles

Aug

men

ted

Rea

lity

(AR

)

Vir

tual

Rea

lity

(VR

)

Gam

ing

Ret

ail

Wea

rabl

eIo

T

Farm

ing

Smar

ten

ergy

Indu

stri

alIn

tern

et

Low Latency Optimize to process a very high volume ofdata messages with minimal delay

X X X X X X X X X X

Increased Bandwidth Ability move a large set amount of datarapidly

X X X X X X X X X X X X

Content Awareness Adaptation of network characteristics ac-cording the local services requirements

X X X X X X X X X X

Low power devices Support for low power devices which haslimited transmission powers

X X X X X X X

Fixed wireless support Operation of wireless systems used to con-nect two fixed locations with a wireless link

X X X X X X X X X X

Fast inter-RAT handoff Speed up the handover takes place betweendifferent RATs

X X X X X X X X

Caching Keeping frequently accessed information ina location close to the requester

X X X X X X X

Edge Analytics An automated analytical computation is per-formed on data at a sensor network switchor other device instead of waiting for thedata to be sent back to a centralized datastore

X X X X X X X X X

Application virtualizationbetween edge and cloud

On demand application and service migra-tion from centralized cloud to the edgecloud

X X X X X X X X X X X

Private or local network Limit the communication and data ex-changes to a certain network segment

X X X X X X X X X X X X

Security Provide localized security X X X X X X XPrivacy Provide localized Privacy X X X X X X XFast Mobility Enable the ability to move or be moved fast

within the network or network coverablearea

X X X X X X X X

B Healthcare

Mobile health and telemedicine are identified as importantuse cases of 5G Wearable low power IoT medical sensorsfor monitoring health related data and tracking records arenow popular in public healthcare facilities [46] Although IoTtechnologies are widely adopted in the health sector [47]their performance goals will not be achievable without edgecomputing solutions like MEC [37] [48] [49] For instancehumanoid robots sitting next to an elderly person may needtactile feedback in 1ms latency for his or her care taking ser-vices Mission critical use cases like remote surgeries requireultra-low latency uninterrupted communication links andcollaborations among surgeons present in different locationsRemote patient monitoring is another use case which enablesconsultants in major cities to interact with patients residingfar away from the medical facility The frequent updates ofhealth records for an elderly person or someone with a chronicdisease needs to proceed ubiquitously and securely With suchpotential use cases and scenarios the role of MEC in healthand social assistance industries becomes more evident [37]

Some research works have already been published aboutthe cooperation between edge computing and IoT in thehealthcare sector In [50] authors describe a military health-

care service platform based on hierarchical IoT architectureand a semantic edge network model The hierarchical IoTarchitecture can collect the vital health parameters of thesoldiers their weapon status as well as their geographicallocations The control center of the battlefield performs therole of edge component which can process and store largeamount of health data sent over an SDN-based network Thepreliminary network architecture proposed in [51] providesreal-time context-aware collaboration for remote robotic tele-surgeries Big data analytics performed by edge computingare also important in e-Healthcare applications [52] In [53]Rahmani et al introduces the smart gateway concept foran IoT-based remote health monitoring system Here they exploit edge computing nodes to update the centralized cloudbased on the medical data generated by the IoT sensors Theirgeo-distributed network of smart e-Health gateways provideslocal data processing for real-time notification for medicalpractitioners secure and privacy preserved data gathering pa-tientsrsquo mobility network interoperability and energy efficientcommunication

C Autonomous VehiclesIoT Automotive5G is a key enabler of V2X (Vehicle to Everything) concept

which covers Vehicle to Vehicle (V2V) vehicle to infras-

7

tructure vehicle to device vehicle to pedestrian vehicle tohome and vehicle to grid [54] In the context of IoT Au-tomotive V2X requires critical communication infrastructurewhere reliability and ultra low latency are crucial factors [55]Use cases in these categories include autonomous and semi-autonomous driving vehicle maintenance and in vehicle info-tainment In order to operate an efficient and reliable vehicularnetwork several features have to be improved these includereal-time traffic monitoring [56] [57] continuous sensing invehicles [58] [59] support for Infotainment applications [60]and improved security [61] However these features cannot beserved by current mobile networks [62] In this vein upcoming5G mobile systems are expected to offer a higher level of flexi-bility leveraging the emerging technologies related to networksoftwarization [63] In this context V2X combined with MECprovides a viable and cost-effective solution that can acceleratedevelopment of V2X and IoT automotive systems [64]

It is important to improve the performance of RAN tech-nologies to enable IoT automatization MEC will play a vitalrole here also For instance MEC technologies may fulfillthe latency reliability and throughput requirements in V2Xchannel modeling of mmWave communication [65] Moreoverthe placement of the MEC server within the RAN providesflexible network services for the vehicle and to efficientlycontrol the radio network resources [66] It is also possibleto design a time-predicted handover mechanism for vehiclesby leveraging road side information at MEC server in order tomeet the demand for high mobility and reliability in vehicularnetworks [66]

In addition ICN-MEC integration can also tackle existingtechnical challenges such as massive mobility of vehicles scal-ability deployment strategies service orchestration massivedata handling fast big data processing as well as ensuringsecurity and privacy [67]

Unmanned aerial vehicles (UAVs) or drones are anothertype of autonomous vehicles which are capable of sensingits environment and navigating without human inputs UAVuse cases include but not limited to public safety smart agri-culture surveillance and environmental monitoring [68] Inorder to maximize the flight time the UAV battery life shouldbe essentially conserved by minimizing the overhead onboardWhen the required processing power exceeds the availableresources on UAV the application data can be offloaded toMEC Accompanying the advanced RATs MEC will facilitatethe offloading process from UAV due to its expected widedeployment in the network [68]

D Gaming AR and VR

Mixed reality (MR) combines virtual reality (VR) andaugmented reality (AR) technologies thereby enabling humansto interact more naturally with the virtual worlds based ondata aggregated by IoT devices [69] With IoT AR tech-nologies are able to benefit directly from the high end inter-connection of objects that characterizes the IoT environmentthrough which users can extend their interactions from thereal world to the virtual world [2] [70] Convergence ofVR and IoT can occur in many ways such as telepresence

tourism industry smart transportation networks and roboticassisted surgeries Exclusive AR and VR experiences withthe delivery of 360 navigable videos will be offered byenhanced mobile broadband connections with low latencyand high reliability for mission-critical services With present-day network standards this might be impossible to achievehowever with the predicted characteristics of 5G such as20 Gbps peak data rate and 1 ms round-trip over-the-airlatency this becomes more easily achievable As identified byETSI MEC will be an ideal solution for low-latency offloadservices in AR and VR applications that combine computergenerated data with physical reality [71] While operatingVR devices over wireless links and deploying the VR controlcenter at MEC server the tracking accuracy can be increasedwith round trip latency of 1 ms and high reliability [72]Migrating computationally intensive tasks to edge servers willincrease the computational capacity of VR devices and savetheir battery-life Furthermore MEC will allow VR devices toaccess cloud resources in an on-demand fashion [73]

MEC platforms provide high capacity and low latencywireless coverage for large venues like stadiums or smartcities with a massive density of users to enjoy the AR andVR experience For instance inside a smart building with anetwork of cameras obtaining raw video frames and preparingthe processed frames for display can be performed locallywith the help of edge computing Furthermore tracking thelocal position of the user or object building a model of theenvironment and identifying known objects in the environ-ment can be offloaded to the edge cloud Similarly in orderto get absolute experience of VR glasses the response timeshould be extremely low When the user moves his head hemay experience delay if the glasses need to access remotedata centers Therefore the expected interaction time betweenmachines and humans needs to be less than 1ms When thelatency of a VR application is more than 1ms the user willexperience cyber sickness which will be interrupting the realVR experience MEC servers in the nearest proximity will beable to serve such applications with ultra low latency Futuregames will be played beyond the entertainment purposes ontop of VR and AR applications which would require theminimum possible latency Pokmon Go and Ingress are twoexamples of successful games that combine AR and sensorinformation such as user location

E Retail

The second largest MEC use case is expected to be in the re-tail businesses [37] Currently IoT has dominated retail marketapplications in many ways including digital signage supplychain management intelligent payment solutions smart vend-ing machines shelves doors resource management stream-ing and safety The high class retail stores which use facialrecognition systems need high definition cameras that generatehuge volumes of data requiring powerful servers within thepremises Therefore the on-site MEC servers will assist toprocess these kind of large data sets produced by IoT devicesin a retail market Big data analytics in shopping centerscan further exploit the collaborative processing between edge

8

and cloud computing [52] Installation of MEC in a retailmarket also provides high speed mobile coverage throughoutthe store WiFi access points that are maintained per store canbe connected to the MEC server to provide WiFi connectivityfor store customers as needed The enabling of MEC willalso omit load balancing Wi-Fi controllers or policy enginesrequired in the wide area networks in the store Althoughnot many academic published research works are explicitlyfocusing on MEC and IoT [74] they have become enormouslyreputed and commercialized technologies in the industry andthe business sectors

F Wearable IoT (WIoT)

During the previous years wearable technology has evolvedtremendously from walkman to step trackers smart watchesto smart glasses The development of low power wirelesstechnologies such as BLE (Bluetooth Low Energy) fuels thedevelopment of wearable devices Present-day wearables spanfrom low-end devices such as health and fitness trackers tohigh-end devices such as VRAR helmets and smart watchesIt is expected that wearables will become the worlds best-selling consumer electronics product after smartphones witha global availability of more than 929 million devices by2021 [75] With the new application domains and enablingservices wearable devices will demand more sophisticatedcommunication infrastructures For instance VRAR wear-ables are demanding gigabits throughput network connectivityto run their applications On the other hand dense deploymentof wearable devices in smart cities will increase the networktraffic on communication networks Thus the next generationcommunication networks should be able to provide the gigabitexperience for the anticipated ultra dense wearable devices[76]

Although cloud computing has enabled wide range of newnetworking services it cannot alone fulfill the upcoming re-quirements for the future wearable ecosystem Mainly the cen-tralized cloud data centers fails due to long End-to-End (E2E)latency Delay-sensitive wearable applications such as VRperceptual stability requires ultra low delay In this contextMEC has the potential to solve the limitations in current cloudbased systems by combining cloud and MEC infrastructuresThis will enable providers deploy storage computing andcaching capabilities in close proximity with such wearabledevices [76]

G IoT in Mechanized Agriculture

In order to meet the demands for future food productionthe agricultural sector will require some major evolution whereIoT will be integrated in various production management andanalytical processes [77] [78] The present-day agriculturalsector has been slow to adopting the emerging Machine-to-Machine (M2M) and IoT technologies when compared withother sectors like smart cities and the medical fields [79]

Precision farming and smart agriculture can be achievedusing autonomous vehicles (tractors) remote monitoring andreal-time analytics It is reported that farmers are increasinglyturning to agricultural drones and satellites to survey their

lands and generate crop data IoT sensors may provide in-formation about crop yields rainfall pest infestation and soilnutrition which are invaluable to production and can improvefarming techniques over time Although low latency is not acritical requirement in smart farming environment manage-ment of large data sets will be a key requirement to considerMEC servers located on-site can assist high tech farmingby collecting and analyzing big data on agriculture in orderto maximize efficiency Likewise without moving everydayfarming applications to a remote cloud MEC platforms canbenefit in terms of data access synchronization storage andother overhead costs the farmer might normally incur

The use of IoT-based automated data collection and moni-toring systems in poultry houses can be used to increase workefficiency and service quality and get a deeper understandingof chicken nurturing [80] Sensing technologies can be usedin carbon dioxide and luminosity sensing these are importantparameters in large scale poultry houses Gas sensors canbe used to get all necessary information to prevent chickeninfertility due to problems such as low carbon dioxide levelsLuminosity senors can help to maintain the proper luminositylevel for optimum productivity Similar to smart farms lowlatency is not a critical requirement in smart poultry houses[80] However it is critical to manage large data sets whereon-site MEC servers can be used In addition sharing the databetween poultry houses and storing legacy data in centralizedservers are important in identifying abnormal incidents in thefarm [81] With the use of MEC poultry houses can workwith intermittent connectivity to the centralized clouds In thatcase MEC servers can temporarily hold the data until farmsare connected with the centralized clouds

H Smart Energy

The smart grid system is an Information CommunicationTechnology (ICT)-enabled energy generation transmissionand distribution network It has capabilities to continuouslysense analyze and monitor both energy flow and energytransportation infrastructure Such features are enabled byadding digital controls and enabling network monitoring andtelecommunication capabilities As a result a smart grid doesnot only provide two-way flows of electrical power but alsoenables real-time automated bidirectional flow of informa-tion Adding such smartness to the aging energy infrastructurewill foster a more efficient energy system

IoT is considered as the foundation for realizing intelli-gence capabilities in smart grid systems IoT integrates theInternet-connectivity into all kinds of grid components suchas transformers breakers switches meters relays intelligentelectronic devices capacitor banks voltage regulators cam-eras and many more These IoT devices are then used tocapture the data required to enable automations IoT-enabledsmart grids provide several benefits such as reduced capitalexpenditure optimized renewable capacity lowered mainte-nance costs and enhanced customer engagement On one handthe transformation of an electrical grid into a smart systemrequires nearly every device and piece of equipment to havebuilt-in secure interconnected intelligence On the other hand

9

an efficient system is required to manage the generated dataie transferring storing and analyzing such huge amounts ofdata which are collected from these smart devices Thereforecloud computing is a viable solution to these IoT-based smartgrids [90]

Generally smart grids are spanning over large geograph-ical areas They often confront bandwidth bottlenecks andcommunication delays due to poor network connectivity andvast number of devices generating data Thus the traditionalcentralized cloud architecture is not suitable for the domain ofthe smart grid since it relies heavily on centralized processing[91] Many delay sensitive smart grid applications such asfault detection isolation and service restoration or VoltVARoptimization cannot tolerate round trip delay to access cen-tralized cloud systems MEC is identified as the viable cloud

computing option to address these limitations MEC allowsthe computation to be performed closer to the data sourceMoreover the potential attack points for the grid is increasingwith the growth of ubiquitous sensor deployment Everysmart IoT device can be vulnerable to potential attacks MECprovides the opportunity to enforce security mechanism closerto the end devices As such even if an attacker gains accessto an endpoint device the attack gets no further informationbeyond the local network segment since MEC has capabilitiesto notice the intrusion and cease the accessibility [85]

I Industrial Internet

The Industrial Internet of Things (IIoT) also known asIndustry 40 [92] is an application of IoT in the domain ofmanufacturing IIoT incorporates numerous advanced commu-

TABLE VI The reviewed state-of-the-art MEC integration in different IoT applications

Ref Description Smar

tH

ome

Smar

tC

ity

Hea

lthca

re

IoT

Aut

omot

ive

Gam

ing

AR

VR

Ret

ail

Wea

rabl

eIo

T

Smar

tA

gric

ultu

re

Smar

tE

nerg

y

Indu

stri

alIn

tern

et

[40] Preliminary design of deploying MEC server functionalities in a smart hometo realize IoT gateway with direct M2M interaction in LTE networks

X

[41] Introduce Gateway-as-a-Service for heterogeneous IoT devices on top of thevirtualization technologies in edge computing

X X

[44] Propose an autonomic creation of MEC services to enhance QoS of videostreaming in smart cities

X

[50] Propose a semantic edge-based IoT architecture for military health services inbattlefield

X

[51] Provide a conceptual MEC based architecture for mission-critical context awarecollaboration in remote surgeries

X

[53] Describe and implement a smart e-Health gateway at the edge of the networksuitable for ubiquitous healthcare systems

X

[64] Analysis on research and engineering challenges co-existence of cloud edgecomputing and data caching strategies at the edge for vehicular networks

X

[82] Discuss the design aspects for the radio access in 5G V2X X[65] Discuss the benefits of merging MEC and mmWave technologies for 5G

applicationsX X X

[66] Propose a novel MEC-based architecture for future cellular vehicular networks X[67] Discuss the benefits of combining ICN and MEC in the context of connected

vehicle environmentsX

[52] Propose a framework for big data analytics between edge and cloud computingplatforms

X X

[74] Design and implement a fog computing based framework that support sharingand reusing contextual data across services in smart city and retail stores

X X

[83] Present a usecase of MEC for Tactile Internet based 5G gaming application X[84] A demonstration of MEC for Tactile Internet based 5G gaming application X[76] Discuss the role of MEC in 5G WIoT communication and its challenges X X[68] Propose an UAV-based IoT platform for a crowd surveillance use case X X[78] Develop and test a ubiquitous sensor network platform for crop lands automa-

tion maintenance in precision agricultureX

[70] Present a serverless edge computing architecture that enables the offloading ofmobile computation with low latency and high throughput using a mobile ARapplication

X

[85] Discuss the benefits of MEC and edge computing (EC) to enhance the securityof smart grids

X

[86] Present a method to optimize the EC based video streaming schemes forIndustrial IoT

X

[87] Present the use of edge computing to provide elastic resources and services toenable microdatabases architecture for IIoT

X

[88] Propose a fog-based communication architecture for Industry 40 applications X[73] Describe research directions and enablers of wireless interconnected VR

systemsX

[89] Design an optimization framework for VRAR communication via small-cellcooperation

X

10

nication and automation technologies such as M2M commu-nication machine learning and big data analytics to improveintelligence and the connectivity [93] For instance IIoTnetworks can connect all of the employees data and processesfrom the factory floor and forward them to the executiveoffices Thus decision makers or employees can create a fulland accurate view of their manufacturing process by usingIIoT network hence improving their ability to make moreinformed decisions IIoT also helps the exploitation as well asimplementation of new intelligent technologies to acceleratethe innovation and transformation of the factory workforce[92]

Primarily IIoT is seen as a way to improve operationalefficiency However IIoT provides a wide range of otherbenefits such as improving connectivity efficiency scalabilitytime savings as well as cost savings for manufacturing pro-cesses with the maximum use of smart machines [92] [94] Ingeneral these smart machines operate with higher accuracygreater efficiency and constant working capabilities than hu-mans [95] Thus IIoT has great potential for improving qualitycontrol sustainability and overall supply chain efficiency

MEC will play a vital role in enabling future IIoT applica-tions [96] by addressing the shortcomings of M2M communi-cation (eg latency resilience cost peer-to-peer connectivitysecurity) in IIoT domain [97] [98] Current market trendsalready show that edge computing will represent many im-plementation scenarios for IIoT For instance real-time edgeanalytics and enhanced edge security are two key drivers inthe creation of new IIoT deployments Thus the addition ofMEC in IIoT networks will fuel the evolution of IIoT as wellas create new business applications [99]

One way to optimize the use of conventional edge com-puting in video streaming schemes for IIoT is presented in[86] By using machine learning algorithms edge computingcan process the sensor data before transmitting to the cloudThis mitigates against the degradation of service quality of thevideo streaming Aggregation of all the sensor data to a singledata center increases latency and raises performance concernsin IIoT domain In order to solve this issue a microdatabasearchitecture is proposed for the Industrial Internet [87] It holdsthe data close to the industrial processes but also makes itavailable near the applications that can benefit from the dataEdge computing also provides elastic resources and services toenable micro-database architecture [87] A fog-based commu-nication architecture for Industry 40 applications is proposedin [88] This approach will substantially minimize the energyconsumption of the IoT nodes Edge computational capabilitiesare further used to predict future data measurements andreduce the throughput from IoT devices to the control unit

III TECHNICAL ASPECTS OF MEC ENABLED IOT

To realize the MEC exploitation for IoT applications thekey value propositions are mostly seen from the technicalparameters such as scalability communication computationoffloading and resource allocation mobility management se-curity privacy and trust management This section describesthe state-of-the-art of each of these technical parameters hence

giving a clear background against which the benefits of MECcan be envisioned

A Scalability

1) Requirements When it comes to actual deployment ofMEC platform for IoT systems scalability is a key factorto consider The compatibility of MEC servers to multiplenetwork environments is one of the factors that will driveits large scale adoption in future networks [100] The IoTenvironment will consist of hundreds of billions of sensors ac-tuators Radio-Frequency Identification (RFID)-tagged objectssoftware vehicles and embedded systems all interconnectedin a huge network of cyber-physical systems At a utility scaleconsideration these devices will be working in close collab-oration to deliver the expected services in technologies likethe smart grids virtual power plants smart homes intelligenttransportation and smart cities That being said the role ofscalability to the realization of such a hyper-connected IoTenvironment becomes more obvious The IoT environment willrequire a dynamic range of capabilities in the network space ifsuch large numbers of devices are to be supported effectively

2) Related work Currently MEC servers have been con-firmed to be compatible with LTE macro base station (eN-odeB) sites 3G Radio Network Controller (RNC) site multi-Radio Access Technology (RAT) cell aggregation site andat the edge of the core network [2] Such multi-RAT cellaggregation schemes can be implemented indoor or outdoorsettings depending on the requirements This invariably en-ables MEC to be applied to many different possible scenariosThe larger the deployment scenarios for MEC the more therange of capabilities it can handle this also translates to higherscalability for MEC-enable technologies like IoT

Designing an edge cloud network implies that an optimallocation for citing the cloud facility is first determined In[105] authors present a design optimization scheme for theMEC architecture based on link-path formulation supportedby heuristics in order to optimize the computation time forthe scheme In this approach consideration is given to bothusers and VMs mobility Hence an optimal point to installthe MEC server is determined through a tread-off betweeninstallation cost and the quality of service to be deliveredTable VII compares the reviewed state-of-the-art scalabilityfeature in MEC enabled IoT

B Communication

1) Requirements There are three main categories for thecommunication concerns about MEC [106] Wireless accesswhile offloading to the mobile edge host Backhaul accesswhile offloading to a remote cloud server Communicationamong IoT devices mobile edge host and remote cloudservers when they collaboratively execute multiple jobs Thefirst and the second categories are the most renowned on behalfof the MEC servers which are the small scale data centersdeployed by the network operators and can be co-located withthe Wireless Access Points (WAPs) In the IoT supportiveMEC systems the consumer devices may communicate with

11

TABLE VII Comparison of the reviewed state-of-the-art scalability feature in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Addressing Search

[101] Discusses the challenges in searching imposed by the burgeoning field of IoT General IoT X[94] Examines a variety of popular and innovative IoT solutions in terms of context-aware

technology perspectives to serve as a conceptual framework for context-aware productdevelopment and research in the IoT paradigm

General IoT solu-tions

X

[102] Proposes an innovative distributed architecture combining machine-to-machine industry-mature protocols (ie MQTT and CoAP) in an original way to enhance the scalabilityof gateways for the efficient IoT-cloud integration

IoT cloud inte-gration

X X

[103] Studies an implementation of edge computing which exploits transparent computing tobuild scalable IoT platforms using transparent computing

Wearable IoT X

[104] Introduces a lightweight edge gateway for the IoT architecture using container-basedvirtualization techniques

General IoT X

the MEC servers either directly or with the support of neigh-boring devices using Device-to-Device (D2D) communicationFor the third category WAPs enable access to the remote datacenters in the central cloud through backhaul links

In order to reap the maximum advantage of computationoffloading leveraged at the edge servers MEC systems needefficient communication channels Unlike the wired connec-tions in the conventional grid computing and cloud computingthe wireless access links between the mobile devices andcloud computing resources in the edge computing paradigmcan be unstable Sudden service outages may occur with theinterruption of access links The inherent challenges withwireless communication channels like multi-path fading in-terference and spectrum shortage should always be taken intoaccount for the design of MEC systems to seamlessly integratecomputation offloading and radio resource management [32]Moreover both wireless and backhaul access links have lim-ited capacities which should be properly shared among mobiledevices in a similar way as sharing the computing resources ofthe MEC server Hence having a cooperative scheme for thejoint allocation of communication and computation resourcesis important for the successful deployment of MEC [106]Redesigning both communication and networking protocolsto integrate communication infrastructures in MEC and IoTsystems is a challenging task The key focus should be onimproving the computation efficiency with respect to datatransmission

Another major requirement is to maintain interoperabilitywhile addressing heterogeneous communication technologiesthat have to be utilized in IoT and MEC paradigms in 5GThere are plenty of radio technologies that facilitate IoT Low-Power Wide Area Networks (LPWANs) (eg WCDMA LTEnarrowband IoT (NB-IoT) Wi-Fi Bluetooth Zigbee SIGFOXand LoRA) The choice of these LPWAN technologies maycreate trade-offs among signal strength operational rangethroughput and power consumption With the arrival of 5Gthe convergence of these communication technologies needsto be achieved since one network will not be fitting based onthose trade-offs

2) Related work Recently Fog-Radio Access Network (F-RAN) was introduced by Peng et al to consolidate the hetero-geneous networks into a single network architecture with 5Geven though they do not operate in the same bands to gain highspectral and operating and energy efficiency [107] Well known

Cloud Radio Access Network (C-RAN) architecture can per-form cooperative transmission across multiple edge nodes withcentralized cloud computing servers via fronthaul links [108]Although C-RAN provides high spectral efficiencies due tothe enhanced interference management capabilities with thecentralized baseband processing at the cloud it has potentiallylarge latencies F-RAN is proposed for 5G MEC deploymentsas an advanced socially aware mobile networking architectureto provide high spectral efficiency while maintaining highenergy efficiency and low latency [107] [108] Precodingdesign resource block allocation user scheduling and cellassociation are jointly designed for radio resource allocation inF-RANs in order to optimize spectral and energy efficienciesand latency performances [109] In [110] Rimal et al pro-pose a unified Time-Division Multiple Access (TDMA) basedresource management scheme for offloading traffic over Fiber-enabled Wireless (FiWi) access networks

In the envisioned 5G systems and MEC architecture bothbackhaul and wireless access links can be facilitated bymillimeter-Wave (mmW) spectrum [111] The use of mmWspectrum will enable high data rate access to MEC function-alities with low latency On the other hand MEC provideslocal computation power usefully for optimizing the perfor-mance of mmW communications In [112] [113] the authorsaddress the joint optimization of communicationcomputationresources with mmW communication They have taken the ad-vantage of blocking probabilities by considering intermittencyof mmW multi-link communications

An open source LPWAN infrastructure called OpenChirpis discussed in [114] OpenChirp which is developed usingLoRWAN allows multiple users to provision and to managebattery-powered transducers across large areas like campusesindustrial zones or cities As pointed out in [30] [115]SDN plays a vital role in improving MEC type technologiesby removing the technical shortcomings in edge computingimplementations The authors summarize the work performedfor implementing MEC based on NFV and SDN where theSDN controller manages the communication between MECservers which form a data center at the edge Table VIIIsummarizes the reviewed state-of-the-art communication is-sues and solutions in MEC enabled IoT

12

TABLE VIII Comparison of the reviewed state-of-the-art communication issues and solutions in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Commnetworkarchitecture

Commresourceallocation

[109] Performance analysis of radio resource allocation in F-RANs for edge cache andadaptive model selection to improve spectral efficiency and energy efficiency

Low latency andhigh reliability

X

[112][113]

Use of mmWave spectrum for high data rate access to MEC servers and backhaul links Low latency andhigh reliability

X

[114] An open source LPWAN infrastructure which allows multiple users to provision andmanage battery-powered transducers across large areas

LPWANinfrastructure

X

[42] A virtualized edge computing architecture with a proxy VM migration scheme tominimize traffic in the core network

IoT big datastreams

X

[115] Proposed network architecture includes multi-interface wireless access network(eg FiWi) heterogeneous backhauling distributed cloudlets hierarchical structure ofa cloudlet and the SDN based mobile core network

IoT big datastreams

X

[110] A novel unified resource management scheme for Ethernet-based FiWi networks thatjointly allocates bandwidth for transmissions of both conventional broadband traffic andMEC data in a TDMA fashion

Mission-criticalIoT

X

[116] Introduce Mobile-IoT-Federation-as-a-Service (MIFaaS) to enable dynamic cooperationamong privatepublic local clouds of IoT devices at the edge of the cellular infrastructureThe selection of the best configuration of federated IoT cloud platforms are modeledas a coalition formation problem

Cellular IoT X

[117] Allocation of radio resources in a joint LTE and NB-IoT system based of MIFaasparadigm [116] Discovered that in handling high-end IoT data traffic a combinationbetween NB-IoT and LTE is essential in providing the needed high data rate and lowlatency

Mission-criticalIoT

X

[118] Integration of D2D communications into edge computing environment reduce transmis-sion delay and traffic load across the network

Mission-criticalIoT

X

[119] Use the theories of stochastic geometry queueing and parallel computing for provi-sioning and planning MEC networks

Communicationlatency

X

C Computation Offloading and Resource Allocation

1) Requirements Computation offloading is the mostprominent and widely discussed feature of MEC that empow-ers resource-constrained IoT devices with augmented com-putational capabilities [29] [33] This will not only prolongthe battery life of the IoT sensor nodes but also reduceE2E latency needed to run sophisticated applications In thefirst place UE has to decide whether to execute the rela-tively simple tasks locally or offload to the MEC servers(ie task model for binary offloading) [32] Secondly thedecision of computation offloading to the MEC servers can beperformed fully or partially In the partial offloading a subsetof computations is executed locally while the rest is offloadedto the MEC server by considering several factors such asusers or application preferences (eg application buffer state)radio and backhaul connections quality (ie between UEand MEC servers) UE capabilities or cloud capabilities andavailability [29]

The sole objective of the offloading policies need to be theminimization of execution delay Other critical concerns areto define the dependency of offloadable components of theapplications based on their ability to partition data (eg real-time user input has to be processed at UE without offloading)and to predict the execution time of multiple tasks Theexecution order or routines have to be carefully formulatedsince certain outcomes can be the inputs of other tasks Aspointed out in [32] the task models for partial offloading canbe represented by task-call graphs with sequential paralleland general dependencies

Although in MEC computation offloading enables power-ful cloud services at the edge level the insufficient batteryenergy at the tiny IoT devices may incur new challenges In

applications like IoT surveillance or remote asset managementthe nodes are typically hard to reach Those applications mayalso require to offload data more frequently in small chunksby consuming more energy Therefore it is necessary to con-sider not only the trade-off between energy consumption andexecution delay in both full and partial offloading scenarios inMEC but also the trade-off between computation energy andtransmission energy consumption in order to extend batterylife

The joint computation and communication resource alloca-tion should be properly addressed in order to get the maximumutilization of available resources Single MEC server will beallocated for the applications which cannot be partitionedThe resources in multiple MEC servers are allocated for theoffloaded applications that can be split into several parts Whena job arrives at the MEC server if there are enough resourcesthe scheduler has to allocate the VM for further processingIf there are no sufficient computation resources it delegatesthe task to the centralized cloud MEC servers also have toallocate computation and communication resources for userapplication jobs and MEC service jobs User mobility networktopology network scalability and load balancing are someother factors to be considered in order to define fare resourceutilization policies on MEC servers Specifically when IoTgateways share limited bandwidth among multiple IoT deviceswhich can handle video audio or bio-medical signals theallocation of bandwidth will become challenging [120] Thelow power wireless technologies (eg BLE ZigBee lowpower Wi-Fi and LPWAN standards like LoRA or SigFox)used in IoT networks have limited bandwidth When the IoTdevices access the MEC server which is acting as the IoTgateway they have to utilize either of those low-power wireless

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

2

A Role of MEC for IoT

Generally cloud computing enables the outsourcing ofstorage and processing functionalities of IoT data to a thirdparty in order to ease the hazel involved in self-managementand data protection However the centralized nature of con-ventional cloud servers may face several challenges such as thesingle point of failure lack of location awareness reachabilityand latencies associated with typical Wide Area Networks(WANs) On the other hand many IoT applications need tobe served with decentralized systems which need mobilitymanagement geo-distribution location awareness scalabilityand ultra-low latency Mission critical communication IoTuse cases need latency as low as 1 ms and reliability ashigh as 9999 For instance factory automation applicationsmay typically require a reliability of 10minus9 packet loss rateand a latency range of 250 micros to 10 ms [3] Thereforethe conjugation of IoT applications and centralized cloudservers may introduce several limitations and vulnerabilitiesIn addition the rapid growth of IoT devices and big data setsmay also create cumbersome traffic on telecommunicationsnetworks

Edge computing was conceived in a bid to fill the gapbetween the centralized cloud and IoT devices Apart fromMEC there are other edge computing paradigms such as Mo-bile Cloud Computing (MCC) fog computing and cloudletsThey tend to coexist with MEC in many technical contextshence the tendency for a misappropriation of these technolo-gies given that they all have similar origin However thesetechnologies are intrinsically different and each of them comeswith its unique value proposition to both existing and futuremobile networks as summarized in Table I

ETSI has identified IoT as one of the key use cases ofMEC [2] MEC has opened many new frontiers for networkoperators service and content providers to deploy versatileand uninterrupted services on IoT applications MEC and IoTfacilitate each other with mutual advantages MEC empowerstiny IoT devices with significant additional computationalcapabilities through computation offloading Similarly IoTexpands MEC services to all types of smart objects rangingfrom sensors and actuators to smart vehicles As shown inFigure 2 MEC servers can perform as gateway nodes whichcan aggregate and process the small data packets generatedby IoT services before they reach the core network Assummarized in [4] the three key benefits of the collaboration

between IoT and MEC are 1) lowering the amount of trafficpassing through the infrastructure 2) reducing the latencyfor applications and services and 3) scaling network servicesdiversely Among these the most significant is the low latencyintroduced by MEC due the reduced physical and virtualcommunication distance

Fig 2 IoT gateway service scenario [2]

B Paper motivation

At present IoT has become a fairly mature technology Asa result the recent decade has seen a plethora of surveyspublished in multiple research areas on IoT including enablingconcepts [5] visions and challenges [6] technologies [7]standardization [8] architecture [9] security [10] [11] pri-vacy [12] trust [13] Social Internet of Things (SIoT) [14]communication [15] context awareness [16] and future direc-tions [6] [17] Few other papers are focused on the combinedaspects of IoT research and their potential application sce-narios [7] [18]ndash[20] Some of these surveys were publishedduring the time when IoT was more of a visionary paradigmthan a real world platform Many future research possibilitiesdiscussed in those papers have already been achieved andcommercialized with high market values However there is yetto be a sufficient number of publications on MEC technologygiven that is relatively a novel technology which lies at theintersection of mobile cloud computing and wireless commu-nication In Table II we summarize the recently published sur-veys on MEC These articles are focused on MEC taxonomyfuture research directions and more specific MEC attributessuch as communication computation offloading security andvirtualization These studies are quite shallow in addressingthe MEC integration with IoT they are mostly focusing onthe requirements and usability of MEC in IoT applicationsIn this short magazine article [21] the authors discuss the

TABLE I High level comparison of edge computing paradigms

MEC Fog computing Cloudlet MCCInitial promotion ETSI (2014) Cisco (2011) Carnegie Mellon Uni (2013) Aepona (2010)Objective Bring cloud computing capabilities closer to User Equipment (UE)Infrastructure owners Telecom operator Private entities individualsNode location Radio network controller or

macro base stationAny strategic location between end user device and cloud

SW architecture Mobile orchestrator based Fog abstraction layer based Cloudlet agent based Service orientedService accessibility Direct access from the closest UE Via Internet connectionLatency and jitter Low HighContext awareness High Medium Low HighStorage capacity andcomputation power

Limited High

Relevance to IoT High Low

3

examples of MEC deployment with special reference to IoTuse cases

To the best of our knowledge there is not a single surveywhich addresses broader range of areas about MEC and itsinfluence on IoT realization Since both MEC and IoT arevery essential to the realization of 5G it is vital to expresstheir associativity in terms of application scenarios and keytechnical attributes Our goal is to broaden the horizons ofpotential inter-dependencies of MEC and IoT technologies andtheir related applications in future 5G and beyond

Furthermore in our previous survey [4] we discuss the roleof MEC in 5G network edge cloud architecture and orchestra-tion There we do not explicitly address the integration of MECfor the realization of IoT and related applications In additionto MEC integration technologies like SDN NFV and networkslicing discussed in [4] we consider ICN in this work There-fore this survey sets to provide a comprehensive overviewof the state-of-the-art technologies which are required for thecomplementary integration of MEC with IoT In this surveyour contributions manifold into three main categories

1) Providing a comprehensive survey on the exploitationof MEC technology for the realization of different IoTapplications

2) Presenting a holistic overview of related works andthe future research directions in areas of scalabilitycommunication computation offloading resource allo-cation mobility management security privacy and trustmanagement of MEC-IoT integration

3) Providing a concise summary of the state-of-the-artMEC integrating technologies for IoT and relatedprojects

C Paper organization

The rest of the paper is organized as follows Section II sum-marizes the well-known IoT applications that require a note-worthy assistance of MEC like edge computing technologiesSection III is particularly focused on technological aspects ofMEC enabled IoT systems in terms of scalability communica-tion computation offloading resource management mobilitymanagement security privacy and trust management Eachtechnical aspect is described with its requirements and relatedworks Section IV and V respectively summarize the relatedwork on different MEC integration technologies and the pro-ceeding research projects in the respective areas Section VIdescribes the lessons learned and the future research directions

TABLE II Summary of important surveys on MEC

Aspect Ref Main contribution Relevance to IoT

Research directions

[22] An elaboration of edge-centric vision and its future re-search challenges

No explicit focus on IoT

[23] A comprehensive overview on sate-of-the-art and futureresearch directions for MEC

Concisely describes how MEC can improve latencyand support big data handling in different IoT deploy-ments

[24] A presentation of MEC related definitions applicationsopportunities and research challenges

Provides no detailed description on IoT Identifies IoTdata handling as a key use case of MEC

[25] A concise tutorial of three edge computing technologiesincluding MEC cloudlets and fog computing

Describes the exploitation of edge computing tech-nologies for IoT with respect to standardization effortsprinciples architectures and applications

[26] A comprehensive survey of relevant research and techno-logical developments in the area of MEC

Identifies MEC services for IoT big-data analytics

Taxonomy [27] A taxonomy of MEC based on different aspects includingits characteristics access technologies applications andobjectives

Classifies MEC applications as computational offload-ing collaborative computing memory replication inIoT and content delivery

[28] A classification of applications deployed in MEC systems No explicit focus on IoTArchitecture andComputationOffloading

[29] A detailed study on decision on computation offloadingallocation of computing resources and mobility man-agement along with a summary of MEC use cases andstandardization efforts

Describes MEC acting as an IoT gateway

Virtualization[4] A survey of 5G network edge cloud architecture and

orchestration with a summary of MEC virtualization tech-nologies including Virtual Machines (VMs) SDN NFVand network slicing

Explains how MEC platform can encompass a localIoT gateway functionality capable of performing dataaggregation and big data analytics for applicationdomains

[30] An investigation on how to exploit SDN for enabling edgecomputing

Discuses SDN scenarios based on IoT and edge Com-puting and the future research

[31] An elaboration of network slicing from an E2E perspectiveon principles enabling technologies and solutions

Describes the role of massive IoT as a key use caseof 5G and network slicing

Communication [32] An comprehensive survey on joint radio-and-computational resource management in MEC systems

Briefly introduces the role of MEC in IoT

[33] A comprehensive survey of issues on computing cachingand communication techniques in MEC

Describes specific applications and use cases of MECin IoT including healthcare wireless sensor systemssmart grid smart home and smart city

MEC-IoT [21] An overview about the role of MEC in IoT use cases Provides examples of MEC deployments for IoT casesSecurity safety and data analytics Vehicle to infras-tructure communication Computation offloading toedge cloud

Security [34] A discussion of the security threats and challenges in theedge paradigms along with the promising solution foreach specific challenge

No explicit discussion on IoT Briefly discusses howIoT will benefit from edge computing and relatedsecurity threats

4

TABLE III Summary of important acronyms

Acronym Definition Acronym Definition

3GPP Third Generation Partnership Project 5G Fifth Generation Wireless NetworkAI Artifical Intelligence AR Augmented RealityBLE Bluetooth Low Energy CaPC Cloud-aware Power ControlCPS Cyber Physical System C-RAN Cloud Radio Access NetworkD2D Device-to-device DDoS Distributed Denial of ServiceDoS Denial of Service E2E End-to-endEC Edge Computing eMBB enhance Mobile BroadbandEMM Energy-aware Mobility Management eNodeB Evolved Node BETSI European Telecommunications Standards Institute EU European UnionFiWi Fiber-enable Wireless F-RAN Fog Radio Access NetworkGDPR General Data Protection Regulation ICN Information Centric NetworkingICT Information Communication Technology IIoT Industrial Internet of ThingsIoT Internet of Things ISG Industry Specification GroupKDN Knowledge-Defined Networking LPWAN Low-power Wide Area NetworkLTE Long Term Evolution M2M Machine-to-machineMANO Management and Orchestration MCC Mobile Cloud ComputingMEC Multi-Access Edge Computing MIFaaS Mobile-IoT-Federation-as-a-ServiceMitM Man-in-the-Middle mmW millimeter-WaveMR Mixed Reality NB-IoT Narrow-band IoTNFV Network Function Virtualization PbD Privacy by DesignQoE Quality of Experience QoS Quality of ServiceRAN Radio Access Networks RAT Radio Access TechnologyRFID Radio-Frequency Identification RNC Radio Network ControllerSCeNB Small Cell eNodeBs SDLB Software Load BalancerSDN Software Defined Networking SDP Software Defined PrivacySIoT Social Internet of Things TDMA Time-division Multiple AccessUAV Unmanned Aerial Vehicles UE User EquipmentV2V Vehicle to Vehicle V2X Vehicle to EverythingVANET Vehicular Ad-hoc Network VM Virtual MachineVNF Virtual Network Function VR Virtual RealityVRARA Virtual RealityAugmented Reality Association WAN Wide Area NetworkingWAP Wireless Access Point WIoT Wearable Internet of ThingsWLAN Wirless Local Area Networking WSN Wireless Sensor Network

Finally Section VII concludes the paper We provide thedefinitions of frequently used acronyms in Table III

II IOT AND MEC APPLICATION SCENARIOS

This section focuses on how IoT can leverage MEC tech-nology in various application scenarios IoT itself is a classicapplication of MEC where the key value proposition of MECis exemplified in a variety of application scenarios (Figure 3)These values become evident in the utility factor measured bythe end user experience while using such IoT related services

Table IV and V respectively show the characteristics of dif-ferent IoT applications and how each application benefits fromMEC-IoT integration In addition Table VI summarizes thereviewed state-of-the-art applications in MEC-IoT domains

A Smart home and Smart city

One of the pioneering applications of the IoT technologyhas been in the areas of home automation and consumerelectronics [39] Several smart home applications that are builton the basis of IoT concept are already available in mostconsumer markets These range from the simple thermostatsensors to other more sophisticated automation systems likesmart metering smart heating and lighting cleaning servicesand home entertainment systems That notwithstanding theamount of data that would be generated on a typical IoTnetwork like the smart home is expected to be huge Hencetransferring such data to the centralized cloud servers will be

impractical with most pre-MEC techniques As a solutionMEC leverages specialized and reliable local services forprocessing and storage capabilities for the large IoT trafficcreated within a building The conventional gateways whichallow IoT applications to run on the centralized cloud can beempowered with MEC-server functionalities [40] [41] Thisextends gateway functionalities to the edge of the networkwith reduced communication latency Since such appliancesare statically deployed in smart home or smart building envi-ronments the cooperation with MEC servers will offer someother features such as easy instantiation relocation privacypreservation and upgrading when necessary [21] [42]

Correspondingly IoT technology has advanced from hometo community and even city scale applications We see nu-merous future promises for public safety health care utilitytourism and the transport sectors Enormous IoT data trafficproduced in smart cities can be ideally processed at the edgeof the network providing low latency and location aware-ness [43] [44] In particular a video cameras (ie deployedfor surveillance) connected with a Long Term Evolution (LTE)network can convey video streams to the MEC server forreal-time processing and anomaly detection [21] Collaborativeedge paradigms that connect multiple MEC servers (iededicated for different services) will advocate the applicationswhich need to process geographically distributed data Forinstance a connected health care application requires to col-laborate with entities from multiple domains such as hospitalpharmacy insurance logistics and government [45]

5

Fig 3 IoT and MEC application scenarios

TABLE IV Characteristics of Different IoT application

IoT Application Data type Data Capacity BackhaulConnectivity

Expected latency Number of IoT Devices

Smart home Stream Historical data

ge 10 MB of data per house-hold per day

Realtime 1 ms -1000 s ge10-100 per house

Smart city Stream Massive data

ge10-100 million GB of dataper city per day

Realtime le1ms ge1000-1million per city

Remote surgery [35] Stream data ge15 million per year Realtime le200 ms ge10-100 per surgeryRemote consultancy Stream data ge 500 million visits per year Realtime 1 ms-100 s 1-10 per appointmentAutonomous vehicles Stream

Massive datage 100 GB per vehicle per day Realtime le1 ms 50-200 per vehicle

AR [36] Stream Massive data

ge1 GBps Realtime le1 ms ge02 million globally

VR [36] Stream Massive data

ge1 GBps Realtime le1ms ge02 million globally

Gaming [36] Stream Massive data

ge10 Mbps Realtime le10 ms ge1 billion globally

Retail [37] Stream Historical data

100 Mbps - 1 Gbps RealtimeIntermittent

le1 ms ge100-1000 per shop

WIoT Stream data lt 1 GB per device Intermittent Several Hours ge1-10 per personFarming Historical data ge 1 GB per farm Intermittent Several hours 100-100000 per farmSmart energy Stream

Massive datage 100000 GB per day Realtime

Intermittent1ms - 10 mins ge 1 billion per grid

Industrial Internet [38] Stream Massive data

ge 100000 GB per day Realtime le1 ms ge 1 million per factory

6

TABLE V MEC and IoT benefits for each application

Required characteristicsof MEC and IoT

Description Smar

tho

me

Smar

tci

ty

Rem

ote

surg

ery

Rem

ote

heal

thco

nsul

tanc

y

Aut

onom

ous

vehi

cles

Aug

men

ted

Rea

lity

(AR

)

Vir

tual

Rea

lity

(VR

)

Gam

ing

Ret

ail

Wea

rabl

eIo

T

Farm

ing

Smar

ten

ergy

Indu

stri

alIn

tern

et

Low Latency Optimize to process a very high volume ofdata messages with minimal delay

X X X X X X X X X X

Increased Bandwidth Ability move a large set amount of datarapidly

X X X X X X X X X X X X

Content Awareness Adaptation of network characteristics ac-cording the local services requirements

X X X X X X X X X X

Low power devices Support for low power devices which haslimited transmission powers

X X X X X X X

Fixed wireless support Operation of wireless systems used to con-nect two fixed locations with a wireless link

X X X X X X X X X X

Fast inter-RAT handoff Speed up the handover takes place betweendifferent RATs

X X X X X X X X

Caching Keeping frequently accessed information ina location close to the requester

X X X X X X X

Edge Analytics An automated analytical computation is per-formed on data at a sensor network switchor other device instead of waiting for thedata to be sent back to a centralized datastore

X X X X X X X X X

Application virtualizationbetween edge and cloud

On demand application and service migra-tion from centralized cloud to the edgecloud

X X X X X X X X X X X

Private or local network Limit the communication and data ex-changes to a certain network segment

X X X X X X X X X X X X

Security Provide localized security X X X X X X XPrivacy Provide localized Privacy X X X X X X XFast Mobility Enable the ability to move or be moved fast

within the network or network coverablearea

X X X X X X X X

B Healthcare

Mobile health and telemedicine are identified as importantuse cases of 5G Wearable low power IoT medical sensorsfor monitoring health related data and tracking records arenow popular in public healthcare facilities [46] Although IoTtechnologies are widely adopted in the health sector [47]their performance goals will not be achievable without edgecomputing solutions like MEC [37] [48] [49] For instancehumanoid robots sitting next to an elderly person may needtactile feedback in 1ms latency for his or her care taking ser-vices Mission critical use cases like remote surgeries requireultra-low latency uninterrupted communication links andcollaborations among surgeons present in different locationsRemote patient monitoring is another use case which enablesconsultants in major cities to interact with patients residingfar away from the medical facility The frequent updates ofhealth records for an elderly person or someone with a chronicdisease needs to proceed ubiquitously and securely With suchpotential use cases and scenarios the role of MEC in healthand social assistance industries becomes more evident [37]

Some research works have already been published aboutthe cooperation between edge computing and IoT in thehealthcare sector In [50] authors describe a military health-

care service platform based on hierarchical IoT architectureand a semantic edge network model The hierarchical IoTarchitecture can collect the vital health parameters of thesoldiers their weapon status as well as their geographicallocations The control center of the battlefield performs therole of edge component which can process and store largeamount of health data sent over an SDN-based network Thepreliminary network architecture proposed in [51] providesreal-time context-aware collaboration for remote robotic tele-surgeries Big data analytics performed by edge computingare also important in e-Healthcare applications [52] In [53]Rahmani et al introduces the smart gateway concept foran IoT-based remote health monitoring system Here they exploit edge computing nodes to update the centralized cloudbased on the medical data generated by the IoT sensors Theirgeo-distributed network of smart e-Health gateways provideslocal data processing for real-time notification for medicalpractitioners secure and privacy preserved data gathering pa-tientsrsquo mobility network interoperability and energy efficientcommunication

C Autonomous VehiclesIoT Automotive5G is a key enabler of V2X (Vehicle to Everything) concept

which covers Vehicle to Vehicle (V2V) vehicle to infras-

7

tructure vehicle to device vehicle to pedestrian vehicle tohome and vehicle to grid [54] In the context of IoT Au-tomotive V2X requires critical communication infrastructurewhere reliability and ultra low latency are crucial factors [55]Use cases in these categories include autonomous and semi-autonomous driving vehicle maintenance and in vehicle info-tainment In order to operate an efficient and reliable vehicularnetwork several features have to be improved these includereal-time traffic monitoring [56] [57] continuous sensing invehicles [58] [59] support for Infotainment applications [60]and improved security [61] However these features cannot beserved by current mobile networks [62] In this vein upcoming5G mobile systems are expected to offer a higher level of flexi-bility leveraging the emerging technologies related to networksoftwarization [63] In this context V2X combined with MECprovides a viable and cost-effective solution that can acceleratedevelopment of V2X and IoT automotive systems [64]

It is important to improve the performance of RAN tech-nologies to enable IoT automatization MEC will play a vitalrole here also For instance MEC technologies may fulfillthe latency reliability and throughput requirements in V2Xchannel modeling of mmWave communication [65] Moreoverthe placement of the MEC server within the RAN providesflexible network services for the vehicle and to efficientlycontrol the radio network resources [66] It is also possibleto design a time-predicted handover mechanism for vehiclesby leveraging road side information at MEC server in order tomeet the demand for high mobility and reliability in vehicularnetworks [66]

In addition ICN-MEC integration can also tackle existingtechnical challenges such as massive mobility of vehicles scal-ability deployment strategies service orchestration massivedata handling fast big data processing as well as ensuringsecurity and privacy [67]

Unmanned aerial vehicles (UAVs) or drones are anothertype of autonomous vehicles which are capable of sensingits environment and navigating without human inputs UAVuse cases include but not limited to public safety smart agri-culture surveillance and environmental monitoring [68] Inorder to maximize the flight time the UAV battery life shouldbe essentially conserved by minimizing the overhead onboardWhen the required processing power exceeds the availableresources on UAV the application data can be offloaded toMEC Accompanying the advanced RATs MEC will facilitatethe offloading process from UAV due to its expected widedeployment in the network [68]

D Gaming AR and VR

Mixed reality (MR) combines virtual reality (VR) andaugmented reality (AR) technologies thereby enabling humansto interact more naturally with the virtual worlds based ondata aggregated by IoT devices [69] With IoT AR tech-nologies are able to benefit directly from the high end inter-connection of objects that characterizes the IoT environmentthrough which users can extend their interactions from thereal world to the virtual world [2] [70] Convergence ofVR and IoT can occur in many ways such as telepresence

tourism industry smart transportation networks and roboticassisted surgeries Exclusive AR and VR experiences withthe delivery of 360 navigable videos will be offered byenhanced mobile broadband connections with low latencyand high reliability for mission-critical services With present-day network standards this might be impossible to achievehowever with the predicted characteristics of 5G such as20 Gbps peak data rate and 1 ms round-trip over-the-airlatency this becomes more easily achievable As identified byETSI MEC will be an ideal solution for low-latency offloadservices in AR and VR applications that combine computergenerated data with physical reality [71] While operatingVR devices over wireless links and deploying the VR controlcenter at MEC server the tracking accuracy can be increasedwith round trip latency of 1 ms and high reliability [72]Migrating computationally intensive tasks to edge servers willincrease the computational capacity of VR devices and savetheir battery-life Furthermore MEC will allow VR devices toaccess cloud resources in an on-demand fashion [73]

MEC platforms provide high capacity and low latencywireless coverage for large venues like stadiums or smartcities with a massive density of users to enjoy the AR andVR experience For instance inside a smart building with anetwork of cameras obtaining raw video frames and preparingthe processed frames for display can be performed locallywith the help of edge computing Furthermore tracking thelocal position of the user or object building a model of theenvironment and identifying known objects in the environ-ment can be offloaded to the edge cloud Similarly in orderto get absolute experience of VR glasses the response timeshould be extremely low When the user moves his head hemay experience delay if the glasses need to access remotedata centers Therefore the expected interaction time betweenmachines and humans needs to be less than 1ms When thelatency of a VR application is more than 1ms the user willexperience cyber sickness which will be interrupting the realVR experience MEC servers in the nearest proximity will beable to serve such applications with ultra low latency Futuregames will be played beyond the entertainment purposes ontop of VR and AR applications which would require theminimum possible latency Pokmon Go and Ingress are twoexamples of successful games that combine AR and sensorinformation such as user location

E Retail

The second largest MEC use case is expected to be in the re-tail businesses [37] Currently IoT has dominated retail marketapplications in many ways including digital signage supplychain management intelligent payment solutions smart vend-ing machines shelves doors resource management stream-ing and safety The high class retail stores which use facialrecognition systems need high definition cameras that generatehuge volumes of data requiring powerful servers within thepremises Therefore the on-site MEC servers will assist toprocess these kind of large data sets produced by IoT devicesin a retail market Big data analytics in shopping centerscan further exploit the collaborative processing between edge

8

and cloud computing [52] Installation of MEC in a retailmarket also provides high speed mobile coverage throughoutthe store WiFi access points that are maintained per store canbe connected to the MEC server to provide WiFi connectivityfor store customers as needed The enabling of MEC willalso omit load balancing Wi-Fi controllers or policy enginesrequired in the wide area networks in the store Althoughnot many academic published research works are explicitlyfocusing on MEC and IoT [74] they have become enormouslyreputed and commercialized technologies in the industry andthe business sectors

F Wearable IoT (WIoT)

During the previous years wearable technology has evolvedtremendously from walkman to step trackers smart watchesto smart glasses The development of low power wirelesstechnologies such as BLE (Bluetooth Low Energy) fuels thedevelopment of wearable devices Present-day wearables spanfrom low-end devices such as health and fitness trackers tohigh-end devices such as VRAR helmets and smart watchesIt is expected that wearables will become the worlds best-selling consumer electronics product after smartphones witha global availability of more than 929 million devices by2021 [75] With the new application domains and enablingservices wearable devices will demand more sophisticatedcommunication infrastructures For instance VRAR wear-ables are demanding gigabits throughput network connectivityto run their applications On the other hand dense deploymentof wearable devices in smart cities will increase the networktraffic on communication networks Thus the next generationcommunication networks should be able to provide the gigabitexperience for the anticipated ultra dense wearable devices[76]

Although cloud computing has enabled wide range of newnetworking services it cannot alone fulfill the upcoming re-quirements for the future wearable ecosystem Mainly the cen-tralized cloud data centers fails due to long End-to-End (E2E)latency Delay-sensitive wearable applications such as VRperceptual stability requires ultra low delay In this contextMEC has the potential to solve the limitations in current cloudbased systems by combining cloud and MEC infrastructuresThis will enable providers deploy storage computing andcaching capabilities in close proximity with such wearabledevices [76]

G IoT in Mechanized Agriculture

In order to meet the demands for future food productionthe agricultural sector will require some major evolution whereIoT will be integrated in various production management andanalytical processes [77] [78] The present-day agriculturalsector has been slow to adopting the emerging Machine-to-Machine (M2M) and IoT technologies when compared withother sectors like smart cities and the medical fields [79]

Precision farming and smart agriculture can be achievedusing autonomous vehicles (tractors) remote monitoring andreal-time analytics It is reported that farmers are increasinglyturning to agricultural drones and satellites to survey their

lands and generate crop data IoT sensors may provide in-formation about crop yields rainfall pest infestation and soilnutrition which are invaluable to production and can improvefarming techniques over time Although low latency is not acritical requirement in smart farming environment manage-ment of large data sets will be a key requirement to considerMEC servers located on-site can assist high tech farmingby collecting and analyzing big data on agriculture in orderto maximize efficiency Likewise without moving everydayfarming applications to a remote cloud MEC platforms canbenefit in terms of data access synchronization storage andother overhead costs the farmer might normally incur

The use of IoT-based automated data collection and moni-toring systems in poultry houses can be used to increase workefficiency and service quality and get a deeper understandingof chicken nurturing [80] Sensing technologies can be usedin carbon dioxide and luminosity sensing these are importantparameters in large scale poultry houses Gas sensors canbe used to get all necessary information to prevent chickeninfertility due to problems such as low carbon dioxide levelsLuminosity senors can help to maintain the proper luminositylevel for optimum productivity Similar to smart farms lowlatency is not a critical requirement in smart poultry houses[80] However it is critical to manage large data sets whereon-site MEC servers can be used In addition sharing the databetween poultry houses and storing legacy data in centralizedservers are important in identifying abnormal incidents in thefarm [81] With the use of MEC poultry houses can workwith intermittent connectivity to the centralized clouds In thatcase MEC servers can temporarily hold the data until farmsare connected with the centralized clouds

H Smart Energy

The smart grid system is an Information CommunicationTechnology (ICT)-enabled energy generation transmissionand distribution network It has capabilities to continuouslysense analyze and monitor both energy flow and energytransportation infrastructure Such features are enabled byadding digital controls and enabling network monitoring andtelecommunication capabilities As a result a smart grid doesnot only provide two-way flows of electrical power but alsoenables real-time automated bidirectional flow of informa-tion Adding such smartness to the aging energy infrastructurewill foster a more efficient energy system

IoT is considered as the foundation for realizing intelli-gence capabilities in smart grid systems IoT integrates theInternet-connectivity into all kinds of grid components suchas transformers breakers switches meters relays intelligentelectronic devices capacitor banks voltage regulators cam-eras and many more These IoT devices are then used tocapture the data required to enable automations IoT-enabledsmart grids provide several benefits such as reduced capitalexpenditure optimized renewable capacity lowered mainte-nance costs and enhanced customer engagement On one handthe transformation of an electrical grid into a smart systemrequires nearly every device and piece of equipment to havebuilt-in secure interconnected intelligence On the other hand

9

an efficient system is required to manage the generated dataie transferring storing and analyzing such huge amounts ofdata which are collected from these smart devices Thereforecloud computing is a viable solution to these IoT-based smartgrids [90]

Generally smart grids are spanning over large geograph-ical areas They often confront bandwidth bottlenecks andcommunication delays due to poor network connectivity andvast number of devices generating data Thus the traditionalcentralized cloud architecture is not suitable for the domain ofthe smart grid since it relies heavily on centralized processing[91] Many delay sensitive smart grid applications such asfault detection isolation and service restoration or VoltVARoptimization cannot tolerate round trip delay to access cen-tralized cloud systems MEC is identified as the viable cloud

computing option to address these limitations MEC allowsthe computation to be performed closer to the data sourceMoreover the potential attack points for the grid is increasingwith the growth of ubiquitous sensor deployment Everysmart IoT device can be vulnerable to potential attacks MECprovides the opportunity to enforce security mechanism closerto the end devices As such even if an attacker gains accessto an endpoint device the attack gets no further informationbeyond the local network segment since MEC has capabilitiesto notice the intrusion and cease the accessibility [85]

I Industrial Internet

The Industrial Internet of Things (IIoT) also known asIndustry 40 [92] is an application of IoT in the domain ofmanufacturing IIoT incorporates numerous advanced commu-

TABLE VI The reviewed state-of-the-art MEC integration in different IoT applications

Ref Description Smar

tH

ome

Smar

tC

ity

Hea

lthca

re

IoT

Aut

omot

ive

Gam

ing

AR

VR

Ret

ail

Wea

rabl

eIo

T

Smar

tA

gric

ultu

re

Smar

tE

nerg

y

Indu

stri

alIn

tern

et

[40] Preliminary design of deploying MEC server functionalities in a smart hometo realize IoT gateway with direct M2M interaction in LTE networks

X

[41] Introduce Gateway-as-a-Service for heterogeneous IoT devices on top of thevirtualization technologies in edge computing

X X

[44] Propose an autonomic creation of MEC services to enhance QoS of videostreaming in smart cities

X

[50] Propose a semantic edge-based IoT architecture for military health services inbattlefield

X

[51] Provide a conceptual MEC based architecture for mission-critical context awarecollaboration in remote surgeries

X

[53] Describe and implement a smart e-Health gateway at the edge of the networksuitable for ubiquitous healthcare systems

X

[64] Analysis on research and engineering challenges co-existence of cloud edgecomputing and data caching strategies at the edge for vehicular networks

X

[82] Discuss the design aspects for the radio access in 5G V2X X[65] Discuss the benefits of merging MEC and mmWave technologies for 5G

applicationsX X X

[66] Propose a novel MEC-based architecture for future cellular vehicular networks X[67] Discuss the benefits of combining ICN and MEC in the context of connected

vehicle environmentsX

[52] Propose a framework for big data analytics between edge and cloud computingplatforms

X X

[74] Design and implement a fog computing based framework that support sharingand reusing contextual data across services in smart city and retail stores

X X

[83] Present a usecase of MEC for Tactile Internet based 5G gaming application X[84] A demonstration of MEC for Tactile Internet based 5G gaming application X[76] Discuss the role of MEC in 5G WIoT communication and its challenges X X[68] Propose an UAV-based IoT platform for a crowd surveillance use case X X[78] Develop and test a ubiquitous sensor network platform for crop lands automa-

tion maintenance in precision agricultureX

[70] Present a serverless edge computing architecture that enables the offloading ofmobile computation with low latency and high throughput using a mobile ARapplication

X

[85] Discuss the benefits of MEC and edge computing (EC) to enhance the securityof smart grids

X

[86] Present a method to optimize the EC based video streaming schemes forIndustrial IoT

X

[87] Present the use of edge computing to provide elastic resources and services toenable microdatabases architecture for IIoT

X

[88] Propose a fog-based communication architecture for Industry 40 applications X[73] Describe research directions and enablers of wireless interconnected VR

systemsX

[89] Design an optimization framework for VRAR communication via small-cellcooperation

X

10

nication and automation technologies such as M2M commu-nication machine learning and big data analytics to improveintelligence and the connectivity [93] For instance IIoTnetworks can connect all of the employees data and processesfrom the factory floor and forward them to the executiveoffices Thus decision makers or employees can create a fulland accurate view of their manufacturing process by usingIIoT network hence improving their ability to make moreinformed decisions IIoT also helps the exploitation as well asimplementation of new intelligent technologies to acceleratethe innovation and transformation of the factory workforce[92]

Primarily IIoT is seen as a way to improve operationalefficiency However IIoT provides a wide range of otherbenefits such as improving connectivity efficiency scalabilitytime savings as well as cost savings for manufacturing pro-cesses with the maximum use of smart machines [92] [94] Ingeneral these smart machines operate with higher accuracygreater efficiency and constant working capabilities than hu-mans [95] Thus IIoT has great potential for improving qualitycontrol sustainability and overall supply chain efficiency

MEC will play a vital role in enabling future IIoT applica-tions [96] by addressing the shortcomings of M2M communi-cation (eg latency resilience cost peer-to-peer connectivitysecurity) in IIoT domain [97] [98] Current market trendsalready show that edge computing will represent many im-plementation scenarios for IIoT For instance real-time edgeanalytics and enhanced edge security are two key drivers inthe creation of new IIoT deployments Thus the addition ofMEC in IIoT networks will fuel the evolution of IIoT as wellas create new business applications [99]

One way to optimize the use of conventional edge com-puting in video streaming schemes for IIoT is presented in[86] By using machine learning algorithms edge computingcan process the sensor data before transmitting to the cloudThis mitigates against the degradation of service quality of thevideo streaming Aggregation of all the sensor data to a singledata center increases latency and raises performance concernsin IIoT domain In order to solve this issue a microdatabasearchitecture is proposed for the Industrial Internet [87] It holdsthe data close to the industrial processes but also makes itavailable near the applications that can benefit from the dataEdge computing also provides elastic resources and services toenable micro-database architecture [87] A fog-based commu-nication architecture for Industry 40 applications is proposedin [88] This approach will substantially minimize the energyconsumption of the IoT nodes Edge computational capabilitiesare further used to predict future data measurements andreduce the throughput from IoT devices to the control unit

III TECHNICAL ASPECTS OF MEC ENABLED IOT

To realize the MEC exploitation for IoT applications thekey value propositions are mostly seen from the technicalparameters such as scalability communication computationoffloading and resource allocation mobility management se-curity privacy and trust management This section describesthe state-of-the-art of each of these technical parameters hence

giving a clear background against which the benefits of MECcan be envisioned

A Scalability

1) Requirements When it comes to actual deployment ofMEC platform for IoT systems scalability is a key factorto consider The compatibility of MEC servers to multiplenetwork environments is one of the factors that will driveits large scale adoption in future networks [100] The IoTenvironment will consist of hundreds of billions of sensors ac-tuators Radio-Frequency Identification (RFID)-tagged objectssoftware vehicles and embedded systems all interconnectedin a huge network of cyber-physical systems At a utility scaleconsideration these devices will be working in close collab-oration to deliver the expected services in technologies likethe smart grids virtual power plants smart homes intelligenttransportation and smart cities That being said the role ofscalability to the realization of such a hyper-connected IoTenvironment becomes more obvious The IoT environment willrequire a dynamic range of capabilities in the network space ifsuch large numbers of devices are to be supported effectively

2) Related work Currently MEC servers have been con-firmed to be compatible with LTE macro base station (eN-odeB) sites 3G Radio Network Controller (RNC) site multi-Radio Access Technology (RAT) cell aggregation site andat the edge of the core network [2] Such multi-RAT cellaggregation schemes can be implemented indoor or outdoorsettings depending on the requirements This invariably en-ables MEC to be applied to many different possible scenariosThe larger the deployment scenarios for MEC the more therange of capabilities it can handle this also translates to higherscalability for MEC-enable technologies like IoT

Designing an edge cloud network implies that an optimallocation for citing the cloud facility is first determined In[105] authors present a design optimization scheme for theMEC architecture based on link-path formulation supportedby heuristics in order to optimize the computation time forthe scheme In this approach consideration is given to bothusers and VMs mobility Hence an optimal point to installthe MEC server is determined through a tread-off betweeninstallation cost and the quality of service to be deliveredTable VII compares the reviewed state-of-the-art scalabilityfeature in MEC enabled IoT

B Communication

1) Requirements There are three main categories for thecommunication concerns about MEC [106] Wireless accesswhile offloading to the mobile edge host Backhaul accesswhile offloading to a remote cloud server Communicationamong IoT devices mobile edge host and remote cloudservers when they collaboratively execute multiple jobs Thefirst and the second categories are the most renowned on behalfof the MEC servers which are the small scale data centersdeployed by the network operators and can be co-located withthe Wireless Access Points (WAPs) In the IoT supportiveMEC systems the consumer devices may communicate with

11

TABLE VII Comparison of the reviewed state-of-the-art scalability feature in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Addressing Search

[101] Discusses the challenges in searching imposed by the burgeoning field of IoT General IoT X[94] Examines a variety of popular and innovative IoT solutions in terms of context-aware

technology perspectives to serve as a conceptual framework for context-aware productdevelopment and research in the IoT paradigm

General IoT solu-tions

X

[102] Proposes an innovative distributed architecture combining machine-to-machine industry-mature protocols (ie MQTT and CoAP) in an original way to enhance the scalabilityof gateways for the efficient IoT-cloud integration

IoT cloud inte-gration

X X

[103] Studies an implementation of edge computing which exploits transparent computing tobuild scalable IoT platforms using transparent computing

Wearable IoT X

[104] Introduces a lightweight edge gateway for the IoT architecture using container-basedvirtualization techniques

General IoT X

the MEC servers either directly or with the support of neigh-boring devices using Device-to-Device (D2D) communicationFor the third category WAPs enable access to the remote datacenters in the central cloud through backhaul links

In order to reap the maximum advantage of computationoffloading leveraged at the edge servers MEC systems needefficient communication channels Unlike the wired connec-tions in the conventional grid computing and cloud computingthe wireless access links between the mobile devices andcloud computing resources in the edge computing paradigmcan be unstable Sudden service outages may occur with theinterruption of access links The inherent challenges withwireless communication channels like multi-path fading in-terference and spectrum shortage should always be taken intoaccount for the design of MEC systems to seamlessly integratecomputation offloading and radio resource management [32]Moreover both wireless and backhaul access links have lim-ited capacities which should be properly shared among mobiledevices in a similar way as sharing the computing resources ofthe MEC server Hence having a cooperative scheme for thejoint allocation of communication and computation resourcesis important for the successful deployment of MEC [106]Redesigning both communication and networking protocolsto integrate communication infrastructures in MEC and IoTsystems is a challenging task The key focus should be onimproving the computation efficiency with respect to datatransmission

Another major requirement is to maintain interoperabilitywhile addressing heterogeneous communication technologiesthat have to be utilized in IoT and MEC paradigms in 5GThere are plenty of radio technologies that facilitate IoT Low-Power Wide Area Networks (LPWANs) (eg WCDMA LTEnarrowband IoT (NB-IoT) Wi-Fi Bluetooth Zigbee SIGFOXand LoRA) The choice of these LPWAN technologies maycreate trade-offs among signal strength operational rangethroughput and power consumption With the arrival of 5Gthe convergence of these communication technologies needsto be achieved since one network will not be fitting based onthose trade-offs

2) Related work Recently Fog-Radio Access Network (F-RAN) was introduced by Peng et al to consolidate the hetero-geneous networks into a single network architecture with 5Geven though they do not operate in the same bands to gain highspectral and operating and energy efficiency [107] Well known

Cloud Radio Access Network (C-RAN) architecture can per-form cooperative transmission across multiple edge nodes withcentralized cloud computing servers via fronthaul links [108]Although C-RAN provides high spectral efficiencies due tothe enhanced interference management capabilities with thecentralized baseband processing at the cloud it has potentiallylarge latencies F-RAN is proposed for 5G MEC deploymentsas an advanced socially aware mobile networking architectureto provide high spectral efficiency while maintaining highenergy efficiency and low latency [107] [108] Precodingdesign resource block allocation user scheduling and cellassociation are jointly designed for radio resource allocation inF-RANs in order to optimize spectral and energy efficienciesand latency performances [109] In [110] Rimal et al pro-pose a unified Time-Division Multiple Access (TDMA) basedresource management scheme for offloading traffic over Fiber-enabled Wireless (FiWi) access networks

In the envisioned 5G systems and MEC architecture bothbackhaul and wireless access links can be facilitated bymillimeter-Wave (mmW) spectrum [111] The use of mmWspectrum will enable high data rate access to MEC function-alities with low latency On the other hand MEC provideslocal computation power usefully for optimizing the perfor-mance of mmW communications In [112] [113] the authorsaddress the joint optimization of communicationcomputationresources with mmW communication They have taken the ad-vantage of blocking probabilities by considering intermittencyof mmW multi-link communications

An open source LPWAN infrastructure called OpenChirpis discussed in [114] OpenChirp which is developed usingLoRWAN allows multiple users to provision and to managebattery-powered transducers across large areas like campusesindustrial zones or cities As pointed out in [30] [115]SDN plays a vital role in improving MEC type technologiesby removing the technical shortcomings in edge computingimplementations The authors summarize the work performedfor implementing MEC based on NFV and SDN where theSDN controller manages the communication between MECservers which form a data center at the edge Table VIIIsummarizes the reviewed state-of-the-art communication is-sues and solutions in MEC enabled IoT

12

TABLE VIII Comparison of the reviewed state-of-the-art communication issues and solutions in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Commnetworkarchitecture

Commresourceallocation

[109] Performance analysis of radio resource allocation in F-RANs for edge cache andadaptive model selection to improve spectral efficiency and energy efficiency

Low latency andhigh reliability

X

[112][113]

Use of mmWave spectrum for high data rate access to MEC servers and backhaul links Low latency andhigh reliability

X

[114] An open source LPWAN infrastructure which allows multiple users to provision andmanage battery-powered transducers across large areas

LPWANinfrastructure

X

[42] A virtualized edge computing architecture with a proxy VM migration scheme tominimize traffic in the core network

IoT big datastreams

X

[115] Proposed network architecture includes multi-interface wireless access network(eg FiWi) heterogeneous backhauling distributed cloudlets hierarchical structure ofa cloudlet and the SDN based mobile core network

IoT big datastreams

X

[110] A novel unified resource management scheme for Ethernet-based FiWi networks thatjointly allocates bandwidth for transmissions of both conventional broadband traffic andMEC data in a TDMA fashion

Mission-criticalIoT

X

[116] Introduce Mobile-IoT-Federation-as-a-Service (MIFaaS) to enable dynamic cooperationamong privatepublic local clouds of IoT devices at the edge of the cellular infrastructureThe selection of the best configuration of federated IoT cloud platforms are modeledas a coalition formation problem

Cellular IoT X

[117] Allocation of radio resources in a joint LTE and NB-IoT system based of MIFaasparadigm [116] Discovered that in handling high-end IoT data traffic a combinationbetween NB-IoT and LTE is essential in providing the needed high data rate and lowlatency

Mission-criticalIoT

X

[118] Integration of D2D communications into edge computing environment reduce transmis-sion delay and traffic load across the network

Mission-criticalIoT

X

[119] Use the theories of stochastic geometry queueing and parallel computing for provi-sioning and planning MEC networks

Communicationlatency

X

C Computation Offloading and Resource Allocation

1) Requirements Computation offloading is the mostprominent and widely discussed feature of MEC that empow-ers resource-constrained IoT devices with augmented com-putational capabilities [29] [33] This will not only prolongthe battery life of the IoT sensor nodes but also reduceE2E latency needed to run sophisticated applications In thefirst place UE has to decide whether to execute the rela-tively simple tasks locally or offload to the MEC servers(ie task model for binary offloading) [32] Secondly thedecision of computation offloading to the MEC servers can beperformed fully or partially In the partial offloading a subsetof computations is executed locally while the rest is offloadedto the MEC server by considering several factors such asusers or application preferences (eg application buffer state)radio and backhaul connections quality (ie between UEand MEC servers) UE capabilities or cloud capabilities andavailability [29]

The sole objective of the offloading policies need to be theminimization of execution delay Other critical concerns areto define the dependency of offloadable components of theapplications based on their ability to partition data (eg real-time user input has to be processed at UE without offloading)and to predict the execution time of multiple tasks Theexecution order or routines have to be carefully formulatedsince certain outcomes can be the inputs of other tasks Aspointed out in [32] the task models for partial offloading canbe represented by task-call graphs with sequential paralleland general dependencies

Although in MEC computation offloading enables power-ful cloud services at the edge level the insufficient batteryenergy at the tiny IoT devices may incur new challenges In

applications like IoT surveillance or remote asset managementthe nodes are typically hard to reach Those applications mayalso require to offload data more frequently in small chunksby consuming more energy Therefore it is necessary to con-sider not only the trade-off between energy consumption andexecution delay in both full and partial offloading scenarios inMEC but also the trade-off between computation energy andtransmission energy consumption in order to extend batterylife

The joint computation and communication resource alloca-tion should be properly addressed in order to get the maximumutilization of available resources Single MEC server will beallocated for the applications which cannot be partitionedThe resources in multiple MEC servers are allocated for theoffloaded applications that can be split into several parts Whena job arrives at the MEC server if there are enough resourcesthe scheduler has to allocate the VM for further processingIf there are no sufficient computation resources it delegatesthe task to the centralized cloud MEC servers also have toallocate computation and communication resources for userapplication jobs and MEC service jobs User mobility networktopology network scalability and load balancing are someother factors to be considered in order to define fare resourceutilization policies on MEC servers Specifically when IoTgateways share limited bandwidth among multiple IoT deviceswhich can handle video audio or bio-medical signals theallocation of bandwidth will become challenging [120] Thelow power wireless technologies (eg BLE ZigBee lowpower Wi-Fi and LPWAN standards like LoRA or SigFox)used in IoT networks have limited bandwidth When the IoTdevices access the MEC server which is acting as the IoTgateway they have to utilize either of those low-power wireless

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

3

examples of MEC deployment with special reference to IoTuse cases

To the best of our knowledge there is not a single surveywhich addresses broader range of areas about MEC and itsinfluence on IoT realization Since both MEC and IoT arevery essential to the realization of 5G it is vital to expresstheir associativity in terms of application scenarios and keytechnical attributes Our goal is to broaden the horizons ofpotential inter-dependencies of MEC and IoT technologies andtheir related applications in future 5G and beyond

Furthermore in our previous survey [4] we discuss the roleof MEC in 5G network edge cloud architecture and orchestra-tion There we do not explicitly address the integration of MECfor the realization of IoT and related applications In additionto MEC integration technologies like SDN NFV and networkslicing discussed in [4] we consider ICN in this work There-fore this survey sets to provide a comprehensive overviewof the state-of-the-art technologies which are required for thecomplementary integration of MEC with IoT In this surveyour contributions manifold into three main categories

1) Providing a comprehensive survey on the exploitationof MEC technology for the realization of different IoTapplications

2) Presenting a holistic overview of related works andthe future research directions in areas of scalabilitycommunication computation offloading resource allo-cation mobility management security privacy and trustmanagement of MEC-IoT integration

3) Providing a concise summary of the state-of-the-artMEC integrating technologies for IoT and relatedprojects

C Paper organization

The rest of the paper is organized as follows Section II sum-marizes the well-known IoT applications that require a note-worthy assistance of MEC like edge computing technologiesSection III is particularly focused on technological aspects ofMEC enabled IoT systems in terms of scalability communica-tion computation offloading resource management mobilitymanagement security privacy and trust management Eachtechnical aspect is described with its requirements and relatedworks Section IV and V respectively summarize the relatedwork on different MEC integration technologies and the pro-ceeding research projects in the respective areas Section VIdescribes the lessons learned and the future research directions

TABLE II Summary of important surveys on MEC

Aspect Ref Main contribution Relevance to IoT

Research directions

[22] An elaboration of edge-centric vision and its future re-search challenges

No explicit focus on IoT

[23] A comprehensive overview on sate-of-the-art and futureresearch directions for MEC

Concisely describes how MEC can improve latencyand support big data handling in different IoT deploy-ments

[24] A presentation of MEC related definitions applicationsopportunities and research challenges

Provides no detailed description on IoT Identifies IoTdata handling as a key use case of MEC

[25] A concise tutorial of three edge computing technologiesincluding MEC cloudlets and fog computing

Describes the exploitation of edge computing tech-nologies for IoT with respect to standardization effortsprinciples architectures and applications

[26] A comprehensive survey of relevant research and techno-logical developments in the area of MEC

Identifies MEC services for IoT big-data analytics

Taxonomy [27] A taxonomy of MEC based on different aspects includingits characteristics access technologies applications andobjectives

Classifies MEC applications as computational offload-ing collaborative computing memory replication inIoT and content delivery

[28] A classification of applications deployed in MEC systems No explicit focus on IoTArchitecture andComputationOffloading

[29] A detailed study on decision on computation offloadingallocation of computing resources and mobility man-agement along with a summary of MEC use cases andstandardization efforts

Describes MEC acting as an IoT gateway

Virtualization[4] A survey of 5G network edge cloud architecture and

orchestration with a summary of MEC virtualization tech-nologies including Virtual Machines (VMs) SDN NFVand network slicing

Explains how MEC platform can encompass a localIoT gateway functionality capable of performing dataaggregation and big data analytics for applicationdomains

[30] An investigation on how to exploit SDN for enabling edgecomputing

Discuses SDN scenarios based on IoT and edge Com-puting and the future research

[31] An elaboration of network slicing from an E2E perspectiveon principles enabling technologies and solutions

Describes the role of massive IoT as a key use caseof 5G and network slicing

Communication [32] An comprehensive survey on joint radio-and-computational resource management in MEC systems

Briefly introduces the role of MEC in IoT

[33] A comprehensive survey of issues on computing cachingand communication techniques in MEC

Describes specific applications and use cases of MECin IoT including healthcare wireless sensor systemssmart grid smart home and smart city

MEC-IoT [21] An overview about the role of MEC in IoT use cases Provides examples of MEC deployments for IoT casesSecurity safety and data analytics Vehicle to infras-tructure communication Computation offloading toedge cloud

Security [34] A discussion of the security threats and challenges in theedge paradigms along with the promising solution foreach specific challenge

No explicit discussion on IoT Briefly discusses howIoT will benefit from edge computing and relatedsecurity threats

4

TABLE III Summary of important acronyms

Acronym Definition Acronym Definition

3GPP Third Generation Partnership Project 5G Fifth Generation Wireless NetworkAI Artifical Intelligence AR Augmented RealityBLE Bluetooth Low Energy CaPC Cloud-aware Power ControlCPS Cyber Physical System C-RAN Cloud Radio Access NetworkD2D Device-to-device DDoS Distributed Denial of ServiceDoS Denial of Service E2E End-to-endEC Edge Computing eMBB enhance Mobile BroadbandEMM Energy-aware Mobility Management eNodeB Evolved Node BETSI European Telecommunications Standards Institute EU European UnionFiWi Fiber-enable Wireless F-RAN Fog Radio Access NetworkGDPR General Data Protection Regulation ICN Information Centric NetworkingICT Information Communication Technology IIoT Industrial Internet of ThingsIoT Internet of Things ISG Industry Specification GroupKDN Knowledge-Defined Networking LPWAN Low-power Wide Area NetworkLTE Long Term Evolution M2M Machine-to-machineMANO Management and Orchestration MCC Mobile Cloud ComputingMEC Multi-Access Edge Computing MIFaaS Mobile-IoT-Federation-as-a-ServiceMitM Man-in-the-Middle mmW millimeter-WaveMR Mixed Reality NB-IoT Narrow-band IoTNFV Network Function Virtualization PbD Privacy by DesignQoE Quality of Experience QoS Quality of ServiceRAN Radio Access Networks RAT Radio Access TechnologyRFID Radio-Frequency Identification RNC Radio Network ControllerSCeNB Small Cell eNodeBs SDLB Software Load BalancerSDN Software Defined Networking SDP Software Defined PrivacySIoT Social Internet of Things TDMA Time-division Multiple AccessUAV Unmanned Aerial Vehicles UE User EquipmentV2V Vehicle to Vehicle V2X Vehicle to EverythingVANET Vehicular Ad-hoc Network VM Virtual MachineVNF Virtual Network Function VR Virtual RealityVRARA Virtual RealityAugmented Reality Association WAN Wide Area NetworkingWAP Wireless Access Point WIoT Wearable Internet of ThingsWLAN Wirless Local Area Networking WSN Wireless Sensor Network

Finally Section VII concludes the paper We provide thedefinitions of frequently used acronyms in Table III

II IOT AND MEC APPLICATION SCENARIOS

This section focuses on how IoT can leverage MEC tech-nology in various application scenarios IoT itself is a classicapplication of MEC where the key value proposition of MECis exemplified in a variety of application scenarios (Figure 3)These values become evident in the utility factor measured bythe end user experience while using such IoT related services

Table IV and V respectively show the characteristics of dif-ferent IoT applications and how each application benefits fromMEC-IoT integration In addition Table VI summarizes thereviewed state-of-the-art applications in MEC-IoT domains

A Smart home and Smart city

One of the pioneering applications of the IoT technologyhas been in the areas of home automation and consumerelectronics [39] Several smart home applications that are builton the basis of IoT concept are already available in mostconsumer markets These range from the simple thermostatsensors to other more sophisticated automation systems likesmart metering smart heating and lighting cleaning servicesand home entertainment systems That notwithstanding theamount of data that would be generated on a typical IoTnetwork like the smart home is expected to be huge Hencetransferring such data to the centralized cloud servers will be

impractical with most pre-MEC techniques As a solutionMEC leverages specialized and reliable local services forprocessing and storage capabilities for the large IoT trafficcreated within a building The conventional gateways whichallow IoT applications to run on the centralized cloud can beempowered with MEC-server functionalities [40] [41] Thisextends gateway functionalities to the edge of the networkwith reduced communication latency Since such appliancesare statically deployed in smart home or smart building envi-ronments the cooperation with MEC servers will offer someother features such as easy instantiation relocation privacypreservation and upgrading when necessary [21] [42]

Correspondingly IoT technology has advanced from hometo community and even city scale applications We see nu-merous future promises for public safety health care utilitytourism and the transport sectors Enormous IoT data trafficproduced in smart cities can be ideally processed at the edgeof the network providing low latency and location aware-ness [43] [44] In particular a video cameras (ie deployedfor surveillance) connected with a Long Term Evolution (LTE)network can convey video streams to the MEC server forreal-time processing and anomaly detection [21] Collaborativeedge paradigms that connect multiple MEC servers (iededicated for different services) will advocate the applicationswhich need to process geographically distributed data Forinstance a connected health care application requires to col-laborate with entities from multiple domains such as hospitalpharmacy insurance logistics and government [45]

5

Fig 3 IoT and MEC application scenarios

TABLE IV Characteristics of Different IoT application

IoT Application Data type Data Capacity BackhaulConnectivity

Expected latency Number of IoT Devices

Smart home Stream Historical data

ge 10 MB of data per house-hold per day

Realtime 1 ms -1000 s ge10-100 per house

Smart city Stream Massive data

ge10-100 million GB of dataper city per day

Realtime le1ms ge1000-1million per city

Remote surgery [35] Stream data ge15 million per year Realtime le200 ms ge10-100 per surgeryRemote consultancy Stream data ge 500 million visits per year Realtime 1 ms-100 s 1-10 per appointmentAutonomous vehicles Stream

Massive datage 100 GB per vehicle per day Realtime le1 ms 50-200 per vehicle

AR [36] Stream Massive data

ge1 GBps Realtime le1 ms ge02 million globally

VR [36] Stream Massive data

ge1 GBps Realtime le1ms ge02 million globally

Gaming [36] Stream Massive data

ge10 Mbps Realtime le10 ms ge1 billion globally

Retail [37] Stream Historical data

100 Mbps - 1 Gbps RealtimeIntermittent

le1 ms ge100-1000 per shop

WIoT Stream data lt 1 GB per device Intermittent Several Hours ge1-10 per personFarming Historical data ge 1 GB per farm Intermittent Several hours 100-100000 per farmSmart energy Stream

Massive datage 100000 GB per day Realtime

Intermittent1ms - 10 mins ge 1 billion per grid

Industrial Internet [38] Stream Massive data

ge 100000 GB per day Realtime le1 ms ge 1 million per factory

6

TABLE V MEC and IoT benefits for each application

Required characteristicsof MEC and IoT

Description Smar

tho

me

Smar

tci

ty

Rem

ote

surg

ery

Rem

ote

heal

thco

nsul

tanc

y

Aut

onom

ous

vehi

cles

Aug

men

ted

Rea

lity

(AR

)

Vir

tual

Rea

lity

(VR

)

Gam

ing

Ret

ail

Wea

rabl

eIo

T

Farm

ing

Smar

ten

ergy

Indu

stri

alIn

tern

et

Low Latency Optimize to process a very high volume ofdata messages with minimal delay

X X X X X X X X X X

Increased Bandwidth Ability move a large set amount of datarapidly

X X X X X X X X X X X X

Content Awareness Adaptation of network characteristics ac-cording the local services requirements

X X X X X X X X X X

Low power devices Support for low power devices which haslimited transmission powers

X X X X X X X

Fixed wireless support Operation of wireless systems used to con-nect two fixed locations with a wireless link

X X X X X X X X X X

Fast inter-RAT handoff Speed up the handover takes place betweendifferent RATs

X X X X X X X X

Caching Keeping frequently accessed information ina location close to the requester

X X X X X X X

Edge Analytics An automated analytical computation is per-formed on data at a sensor network switchor other device instead of waiting for thedata to be sent back to a centralized datastore

X X X X X X X X X

Application virtualizationbetween edge and cloud

On demand application and service migra-tion from centralized cloud to the edgecloud

X X X X X X X X X X X

Private or local network Limit the communication and data ex-changes to a certain network segment

X X X X X X X X X X X X

Security Provide localized security X X X X X X XPrivacy Provide localized Privacy X X X X X X XFast Mobility Enable the ability to move or be moved fast

within the network or network coverablearea

X X X X X X X X

B Healthcare

Mobile health and telemedicine are identified as importantuse cases of 5G Wearable low power IoT medical sensorsfor monitoring health related data and tracking records arenow popular in public healthcare facilities [46] Although IoTtechnologies are widely adopted in the health sector [47]their performance goals will not be achievable without edgecomputing solutions like MEC [37] [48] [49] For instancehumanoid robots sitting next to an elderly person may needtactile feedback in 1ms latency for his or her care taking ser-vices Mission critical use cases like remote surgeries requireultra-low latency uninterrupted communication links andcollaborations among surgeons present in different locationsRemote patient monitoring is another use case which enablesconsultants in major cities to interact with patients residingfar away from the medical facility The frequent updates ofhealth records for an elderly person or someone with a chronicdisease needs to proceed ubiquitously and securely With suchpotential use cases and scenarios the role of MEC in healthand social assistance industries becomes more evident [37]

Some research works have already been published aboutthe cooperation between edge computing and IoT in thehealthcare sector In [50] authors describe a military health-

care service platform based on hierarchical IoT architectureand a semantic edge network model The hierarchical IoTarchitecture can collect the vital health parameters of thesoldiers their weapon status as well as their geographicallocations The control center of the battlefield performs therole of edge component which can process and store largeamount of health data sent over an SDN-based network Thepreliminary network architecture proposed in [51] providesreal-time context-aware collaboration for remote robotic tele-surgeries Big data analytics performed by edge computingare also important in e-Healthcare applications [52] In [53]Rahmani et al introduces the smart gateway concept foran IoT-based remote health monitoring system Here they exploit edge computing nodes to update the centralized cloudbased on the medical data generated by the IoT sensors Theirgeo-distributed network of smart e-Health gateways provideslocal data processing for real-time notification for medicalpractitioners secure and privacy preserved data gathering pa-tientsrsquo mobility network interoperability and energy efficientcommunication

C Autonomous VehiclesIoT Automotive5G is a key enabler of V2X (Vehicle to Everything) concept

which covers Vehicle to Vehicle (V2V) vehicle to infras-

7

tructure vehicle to device vehicle to pedestrian vehicle tohome and vehicle to grid [54] In the context of IoT Au-tomotive V2X requires critical communication infrastructurewhere reliability and ultra low latency are crucial factors [55]Use cases in these categories include autonomous and semi-autonomous driving vehicle maintenance and in vehicle info-tainment In order to operate an efficient and reliable vehicularnetwork several features have to be improved these includereal-time traffic monitoring [56] [57] continuous sensing invehicles [58] [59] support for Infotainment applications [60]and improved security [61] However these features cannot beserved by current mobile networks [62] In this vein upcoming5G mobile systems are expected to offer a higher level of flexi-bility leveraging the emerging technologies related to networksoftwarization [63] In this context V2X combined with MECprovides a viable and cost-effective solution that can acceleratedevelopment of V2X and IoT automotive systems [64]

It is important to improve the performance of RAN tech-nologies to enable IoT automatization MEC will play a vitalrole here also For instance MEC technologies may fulfillthe latency reliability and throughput requirements in V2Xchannel modeling of mmWave communication [65] Moreoverthe placement of the MEC server within the RAN providesflexible network services for the vehicle and to efficientlycontrol the radio network resources [66] It is also possibleto design a time-predicted handover mechanism for vehiclesby leveraging road side information at MEC server in order tomeet the demand for high mobility and reliability in vehicularnetworks [66]

In addition ICN-MEC integration can also tackle existingtechnical challenges such as massive mobility of vehicles scal-ability deployment strategies service orchestration massivedata handling fast big data processing as well as ensuringsecurity and privacy [67]

Unmanned aerial vehicles (UAVs) or drones are anothertype of autonomous vehicles which are capable of sensingits environment and navigating without human inputs UAVuse cases include but not limited to public safety smart agri-culture surveillance and environmental monitoring [68] Inorder to maximize the flight time the UAV battery life shouldbe essentially conserved by minimizing the overhead onboardWhen the required processing power exceeds the availableresources on UAV the application data can be offloaded toMEC Accompanying the advanced RATs MEC will facilitatethe offloading process from UAV due to its expected widedeployment in the network [68]

D Gaming AR and VR

Mixed reality (MR) combines virtual reality (VR) andaugmented reality (AR) technologies thereby enabling humansto interact more naturally with the virtual worlds based ondata aggregated by IoT devices [69] With IoT AR tech-nologies are able to benefit directly from the high end inter-connection of objects that characterizes the IoT environmentthrough which users can extend their interactions from thereal world to the virtual world [2] [70] Convergence ofVR and IoT can occur in many ways such as telepresence

tourism industry smart transportation networks and roboticassisted surgeries Exclusive AR and VR experiences withthe delivery of 360 navigable videos will be offered byenhanced mobile broadband connections with low latencyand high reliability for mission-critical services With present-day network standards this might be impossible to achievehowever with the predicted characteristics of 5G such as20 Gbps peak data rate and 1 ms round-trip over-the-airlatency this becomes more easily achievable As identified byETSI MEC will be an ideal solution for low-latency offloadservices in AR and VR applications that combine computergenerated data with physical reality [71] While operatingVR devices over wireless links and deploying the VR controlcenter at MEC server the tracking accuracy can be increasedwith round trip latency of 1 ms and high reliability [72]Migrating computationally intensive tasks to edge servers willincrease the computational capacity of VR devices and savetheir battery-life Furthermore MEC will allow VR devices toaccess cloud resources in an on-demand fashion [73]

MEC platforms provide high capacity and low latencywireless coverage for large venues like stadiums or smartcities with a massive density of users to enjoy the AR andVR experience For instance inside a smart building with anetwork of cameras obtaining raw video frames and preparingthe processed frames for display can be performed locallywith the help of edge computing Furthermore tracking thelocal position of the user or object building a model of theenvironment and identifying known objects in the environ-ment can be offloaded to the edge cloud Similarly in orderto get absolute experience of VR glasses the response timeshould be extremely low When the user moves his head hemay experience delay if the glasses need to access remotedata centers Therefore the expected interaction time betweenmachines and humans needs to be less than 1ms When thelatency of a VR application is more than 1ms the user willexperience cyber sickness which will be interrupting the realVR experience MEC servers in the nearest proximity will beable to serve such applications with ultra low latency Futuregames will be played beyond the entertainment purposes ontop of VR and AR applications which would require theminimum possible latency Pokmon Go and Ingress are twoexamples of successful games that combine AR and sensorinformation such as user location

E Retail

The second largest MEC use case is expected to be in the re-tail businesses [37] Currently IoT has dominated retail marketapplications in many ways including digital signage supplychain management intelligent payment solutions smart vend-ing machines shelves doors resource management stream-ing and safety The high class retail stores which use facialrecognition systems need high definition cameras that generatehuge volumes of data requiring powerful servers within thepremises Therefore the on-site MEC servers will assist toprocess these kind of large data sets produced by IoT devicesin a retail market Big data analytics in shopping centerscan further exploit the collaborative processing between edge

8

and cloud computing [52] Installation of MEC in a retailmarket also provides high speed mobile coverage throughoutthe store WiFi access points that are maintained per store canbe connected to the MEC server to provide WiFi connectivityfor store customers as needed The enabling of MEC willalso omit load balancing Wi-Fi controllers or policy enginesrequired in the wide area networks in the store Althoughnot many academic published research works are explicitlyfocusing on MEC and IoT [74] they have become enormouslyreputed and commercialized technologies in the industry andthe business sectors

F Wearable IoT (WIoT)

During the previous years wearable technology has evolvedtremendously from walkman to step trackers smart watchesto smart glasses The development of low power wirelesstechnologies such as BLE (Bluetooth Low Energy) fuels thedevelopment of wearable devices Present-day wearables spanfrom low-end devices such as health and fitness trackers tohigh-end devices such as VRAR helmets and smart watchesIt is expected that wearables will become the worlds best-selling consumer electronics product after smartphones witha global availability of more than 929 million devices by2021 [75] With the new application domains and enablingservices wearable devices will demand more sophisticatedcommunication infrastructures For instance VRAR wear-ables are demanding gigabits throughput network connectivityto run their applications On the other hand dense deploymentof wearable devices in smart cities will increase the networktraffic on communication networks Thus the next generationcommunication networks should be able to provide the gigabitexperience for the anticipated ultra dense wearable devices[76]

Although cloud computing has enabled wide range of newnetworking services it cannot alone fulfill the upcoming re-quirements for the future wearable ecosystem Mainly the cen-tralized cloud data centers fails due to long End-to-End (E2E)latency Delay-sensitive wearable applications such as VRperceptual stability requires ultra low delay In this contextMEC has the potential to solve the limitations in current cloudbased systems by combining cloud and MEC infrastructuresThis will enable providers deploy storage computing andcaching capabilities in close proximity with such wearabledevices [76]

G IoT in Mechanized Agriculture

In order to meet the demands for future food productionthe agricultural sector will require some major evolution whereIoT will be integrated in various production management andanalytical processes [77] [78] The present-day agriculturalsector has been slow to adopting the emerging Machine-to-Machine (M2M) and IoT technologies when compared withother sectors like smart cities and the medical fields [79]

Precision farming and smart agriculture can be achievedusing autonomous vehicles (tractors) remote monitoring andreal-time analytics It is reported that farmers are increasinglyturning to agricultural drones and satellites to survey their

lands and generate crop data IoT sensors may provide in-formation about crop yields rainfall pest infestation and soilnutrition which are invaluable to production and can improvefarming techniques over time Although low latency is not acritical requirement in smart farming environment manage-ment of large data sets will be a key requirement to considerMEC servers located on-site can assist high tech farmingby collecting and analyzing big data on agriculture in orderto maximize efficiency Likewise without moving everydayfarming applications to a remote cloud MEC platforms canbenefit in terms of data access synchronization storage andother overhead costs the farmer might normally incur

The use of IoT-based automated data collection and moni-toring systems in poultry houses can be used to increase workefficiency and service quality and get a deeper understandingof chicken nurturing [80] Sensing technologies can be usedin carbon dioxide and luminosity sensing these are importantparameters in large scale poultry houses Gas sensors canbe used to get all necessary information to prevent chickeninfertility due to problems such as low carbon dioxide levelsLuminosity senors can help to maintain the proper luminositylevel for optimum productivity Similar to smart farms lowlatency is not a critical requirement in smart poultry houses[80] However it is critical to manage large data sets whereon-site MEC servers can be used In addition sharing the databetween poultry houses and storing legacy data in centralizedservers are important in identifying abnormal incidents in thefarm [81] With the use of MEC poultry houses can workwith intermittent connectivity to the centralized clouds In thatcase MEC servers can temporarily hold the data until farmsare connected with the centralized clouds

H Smart Energy

The smart grid system is an Information CommunicationTechnology (ICT)-enabled energy generation transmissionand distribution network It has capabilities to continuouslysense analyze and monitor both energy flow and energytransportation infrastructure Such features are enabled byadding digital controls and enabling network monitoring andtelecommunication capabilities As a result a smart grid doesnot only provide two-way flows of electrical power but alsoenables real-time automated bidirectional flow of informa-tion Adding such smartness to the aging energy infrastructurewill foster a more efficient energy system

IoT is considered as the foundation for realizing intelli-gence capabilities in smart grid systems IoT integrates theInternet-connectivity into all kinds of grid components suchas transformers breakers switches meters relays intelligentelectronic devices capacitor banks voltage regulators cam-eras and many more These IoT devices are then used tocapture the data required to enable automations IoT-enabledsmart grids provide several benefits such as reduced capitalexpenditure optimized renewable capacity lowered mainte-nance costs and enhanced customer engagement On one handthe transformation of an electrical grid into a smart systemrequires nearly every device and piece of equipment to havebuilt-in secure interconnected intelligence On the other hand

9

an efficient system is required to manage the generated dataie transferring storing and analyzing such huge amounts ofdata which are collected from these smart devices Thereforecloud computing is a viable solution to these IoT-based smartgrids [90]

Generally smart grids are spanning over large geograph-ical areas They often confront bandwidth bottlenecks andcommunication delays due to poor network connectivity andvast number of devices generating data Thus the traditionalcentralized cloud architecture is not suitable for the domain ofthe smart grid since it relies heavily on centralized processing[91] Many delay sensitive smart grid applications such asfault detection isolation and service restoration or VoltVARoptimization cannot tolerate round trip delay to access cen-tralized cloud systems MEC is identified as the viable cloud

computing option to address these limitations MEC allowsthe computation to be performed closer to the data sourceMoreover the potential attack points for the grid is increasingwith the growth of ubiquitous sensor deployment Everysmart IoT device can be vulnerable to potential attacks MECprovides the opportunity to enforce security mechanism closerto the end devices As such even if an attacker gains accessto an endpoint device the attack gets no further informationbeyond the local network segment since MEC has capabilitiesto notice the intrusion and cease the accessibility [85]

I Industrial Internet

The Industrial Internet of Things (IIoT) also known asIndustry 40 [92] is an application of IoT in the domain ofmanufacturing IIoT incorporates numerous advanced commu-

TABLE VI The reviewed state-of-the-art MEC integration in different IoT applications

Ref Description Smar

tH

ome

Smar

tC

ity

Hea

lthca

re

IoT

Aut

omot

ive

Gam

ing

AR

VR

Ret

ail

Wea

rabl

eIo

T

Smar

tA

gric

ultu

re

Smar

tE

nerg

y

Indu

stri

alIn

tern

et

[40] Preliminary design of deploying MEC server functionalities in a smart hometo realize IoT gateway with direct M2M interaction in LTE networks

X

[41] Introduce Gateway-as-a-Service for heterogeneous IoT devices on top of thevirtualization technologies in edge computing

X X

[44] Propose an autonomic creation of MEC services to enhance QoS of videostreaming in smart cities

X

[50] Propose a semantic edge-based IoT architecture for military health services inbattlefield

X

[51] Provide a conceptual MEC based architecture for mission-critical context awarecollaboration in remote surgeries

X

[53] Describe and implement a smart e-Health gateway at the edge of the networksuitable for ubiquitous healthcare systems

X

[64] Analysis on research and engineering challenges co-existence of cloud edgecomputing and data caching strategies at the edge for vehicular networks

X

[82] Discuss the design aspects for the radio access in 5G V2X X[65] Discuss the benefits of merging MEC and mmWave technologies for 5G

applicationsX X X

[66] Propose a novel MEC-based architecture for future cellular vehicular networks X[67] Discuss the benefits of combining ICN and MEC in the context of connected

vehicle environmentsX

[52] Propose a framework for big data analytics between edge and cloud computingplatforms

X X

[74] Design and implement a fog computing based framework that support sharingand reusing contextual data across services in smart city and retail stores

X X

[83] Present a usecase of MEC for Tactile Internet based 5G gaming application X[84] A demonstration of MEC for Tactile Internet based 5G gaming application X[76] Discuss the role of MEC in 5G WIoT communication and its challenges X X[68] Propose an UAV-based IoT platform for a crowd surveillance use case X X[78] Develop and test a ubiquitous sensor network platform for crop lands automa-

tion maintenance in precision agricultureX

[70] Present a serverless edge computing architecture that enables the offloading ofmobile computation with low latency and high throughput using a mobile ARapplication

X

[85] Discuss the benefits of MEC and edge computing (EC) to enhance the securityof smart grids

X

[86] Present a method to optimize the EC based video streaming schemes forIndustrial IoT

X

[87] Present the use of edge computing to provide elastic resources and services toenable microdatabases architecture for IIoT

X

[88] Propose a fog-based communication architecture for Industry 40 applications X[73] Describe research directions and enablers of wireless interconnected VR

systemsX

[89] Design an optimization framework for VRAR communication via small-cellcooperation

X

10

nication and automation technologies such as M2M commu-nication machine learning and big data analytics to improveintelligence and the connectivity [93] For instance IIoTnetworks can connect all of the employees data and processesfrom the factory floor and forward them to the executiveoffices Thus decision makers or employees can create a fulland accurate view of their manufacturing process by usingIIoT network hence improving their ability to make moreinformed decisions IIoT also helps the exploitation as well asimplementation of new intelligent technologies to acceleratethe innovation and transformation of the factory workforce[92]

Primarily IIoT is seen as a way to improve operationalefficiency However IIoT provides a wide range of otherbenefits such as improving connectivity efficiency scalabilitytime savings as well as cost savings for manufacturing pro-cesses with the maximum use of smart machines [92] [94] Ingeneral these smart machines operate with higher accuracygreater efficiency and constant working capabilities than hu-mans [95] Thus IIoT has great potential for improving qualitycontrol sustainability and overall supply chain efficiency

MEC will play a vital role in enabling future IIoT applica-tions [96] by addressing the shortcomings of M2M communi-cation (eg latency resilience cost peer-to-peer connectivitysecurity) in IIoT domain [97] [98] Current market trendsalready show that edge computing will represent many im-plementation scenarios for IIoT For instance real-time edgeanalytics and enhanced edge security are two key drivers inthe creation of new IIoT deployments Thus the addition ofMEC in IIoT networks will fuel the evolution of IIoT as wellas create new business applications [99]

One way to optimize the use of conventional edge com-puting in video streaming schemes for IIoT is presented in[86] By using machine learning algorithms edge computingcan process the sensor data before transmitting to the cloudThis mitigates against the degradation of service quality of thevideo streaming Aggregation of all the sensor data to a singledata center increases latency and raises performance concernsin IIoT domain In order to solve this issue a microdatabasearchitecture is proposed for the Industrial Internet [87] It holdsthe data close to the industrial processes but also makes itavailable near the applications that can benefit from the dataEdge computing also provides elastic resources and services toenable micro-database architecture [87] A fog-based commu-nication architecture for Industry 40 applications is proposedin [88] This approach will substantially minimize the energyconsumption of the IoT nodes Edge computational capabilitiesare further used to predict future data measurements andreduce the throughput from IoT devices to the control unit

III TECHNICAL ASPECTS OF MEC ENABLED IOT

To realize the MEC exploitation for IoT applications thekey value propositions are mostly seen from the technicalparameters such as scalability communication computationoffloading and resource allocation mobility management se-curity privacy and trust management This section describesthe state-of-the-art of each of these technical parameters hence

giving a clear background against which the benefits of MECcan be envisioned

A Scalability

1) Requirements When it comes to actual deployment ofMEC platform for IoT systems scalability is a key factorto consider The compatibility of MEC servers to multiplenetwork environments is one of the factors that will driveits large scale adoption in future networks [100] The IoTenvironment will consist of hundreds of billions of sensors ac-tuators Radio-Frequency Identification (RFID)-tagged objectssoftware vehicles and embedded systems all interconnectedin a huge network of cyber-physical systems At a utility scaleconsideration these devices will be working in close collab-oration to deliver the expected services in technologies likethe smart grids virtual power plants smart homes intelligenttransportation and smart cities That being said the role ofscalability to the realization of such a hyper-connected IoTenvironment becomes more obvious The IoT environment willrequire a dynamic range of capabilities in the network space ifsuch large numbers of devices are to be supported effectively

2) Related work Currently MEC servers have been con-firmed to be compatible with LTE macro base station (eN-odeB) sites 3G Radio Network Controller (RNC) site multi-Radio Access Technology (RAT) cell aggregation site andat the edge of the core network [2] Such multi-RAT cellaggregation schemes can be implemented indoor or outdoorsettings depending on the requirements This invariably en-ables MEC to be applied to many different possible scenariosThe larger the deployment scenarios for MEC the more therange of capabilities it can handle this also translates to higherscalability for MEC-enable technologies like IoT

Designing an edge cloud network implies that an optimallocation for citing the cloud facility is first determined In[105] authors present a design optimization scheme for theMEC architecture based on link-path formulation supportedby heuristics in order to optimize the computation time forthe scheme In this approach consideration is given to bothusers and VMs mobility Hence an optimal point to installthe MEC server is determined through a tread-off betweeninstallation cost and the quality of service to be deliveredTable VII compares the reviewed state-of-the-art scalabilityfeature in MEC enabled IoT

B Communication

1) Requirements There are three main categories for thecommunication concerns about MEC [106] Wireless accesswhile offloading to the mobile edge host Backhaul accesswhile offloading to a remote cloud server Communicationamong IoT devices mobile edge host and remote cloudservers when they collaboratively execute multiple jobs Thefirst and the second categories are the most renowned on behalfof the MEC servers which are the small scale data centersdeployed by the network operators and can be co-located withthe Wireless Access Points (WAPs) In the IoT supportiveMEC systems the consumer devices may communicate with

11

TABLE VII Comparison of the reviewed state-of-the-art scalability feature in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Addressing Search

[101] Discusses the challenges in searching imposed by the burgeoning field of IoT General IoT X[94] Examines a variety of popular and innovative IoT solutions in terms of context-aware

technology perspectives to serve as a conceptual framework for context-aware productdevelopment and research in the IoT paradigm

General IoT solu-tions

X

[102] Proposes an innovative distributed architecture combining machine-to-machine industry-mature protocols (ie MQTT and CoAP) in an original way to enhance the scalabilityof gateways for the efficient IoT-cloud integration

IoT cloud inte-gration

X X

[103] Studies an implementation of edge computing which exploits transparent computing tobuild scalable IoT platforms using transparent computing

Wearable IoT X

[104] Introduces a lightweight edge gateway for the IoT architecture using container-basedvirtualization techniques

General IoT X

the MEC servers either directly or with the support of neigh-boring devices using Device-to-Device (D2D) communicationFor the third category WAPs enable access to the remote datacenters in the central cloud through backhaul links

In order to reap the maximum advantage of computationoffloading leveraged at the edge servers MEC systems needefficient communication channels Unlike the wired connec-tions in the conventional grid computing and cloud computingthe wireless access links between the mobile devices andcloud computing resources in the edge computing paradigmcan be unstable Sudden service outages may occur with theinterruption of access links The inherent challenges withwireless communication channels like multi-path fading in-terference and spectrum shortage should always be taken intoaccount for the design of MEC systems to seamlessly integratecomputation offloading and radio resource management [32]Moreover both wireless and backhaul access links have lim-ited capacities which should be properly shared among mobiledevices in a similar way as sharing the computing resources ofthe MEC server Hence having a cooperative scheme for thejoint allocation of communication and computation resourcesis important for the successful deployment of MEC [106]Redesigning both communication and networking protocolsto integrate communication infrastructures in MEC and IoTsystems is a challenging task The key focus should be onimproving the computation efficiency with respect to datatransmission

Another major requirement is to maintain interoperabilitywhile addressing heterogeneous communication technologiesthat have to be utilized in IoT and MEC paradigms in 5GThere are plenty of radio technologies that facilitate IoT Low-Power Wide Area Networks (LPWANs) (eg WCDMA LTEnarrowband IoT (NB-IoT) Wi-Fi Bluetooth Zigbee SIGFOXand LoRA) The choice of these LPWAN technologies maycreate trade-offs among signal strength operational rangethroughput and power consumption With the arrival of 5Gthe convergence of these communication technologies needsto be achieved since one network will not be fitting based onthose trade-offs

2) Related work Recently Fog-Radio Access Network (F-RAN) was introduced by Peng et al to consolidate the hetero-geneous networks into a single network architecture with 5Geven though they do not operate in the same bands to gain highspectral and operating and energy efficiency [107] Well known

Cloud Radio Access Network (C-RAN) architecture can per-form cooperative transmission across multiple edge nodes withcentralized cloud computing servers via fronthaul links [108]Although C-RAN provides high spectral efficiencies due tothe enhanced interference management capabilities with thecentralized baseband processing at the cloud it has potentiallylarge latencies F-RAN is proposed for 5G MEC deploymentsas an advanced socially aware mobile networking architectureto provide high spectral efficiency while maintaining highenergy efficiency and low latency [107] [108] Precodingdesign resource block allocation user scheduling and cellassociation are jointly designed for radio resource allocation inF-RANs in order to optimize spectral and energy efficienciesand latency performances [109] In [110] Rimal et al pro-pose a unified Time-Division Multiple Access (TDMA) basedresource management scheme for offloading traffic over Fiber-enabled Wireless (FiWi) access networks

In the envisioned 5G systems and MEC architecture bothbackhaul and wireless access links can be facilitated bymillimeter-Wave (mmW) spectrum [111] The use of mmWspectrum will enable high data rate access to MEC function-alities with low latency On the other hand MEC provideslocal computation power usefully for optimizing the perfor-mance of mmW communications In [112] [113] the authorsaddress the joint optimization of communicationcomputationresources with mmW communication They have taken the ad-vantage of blocking probabilities by considering intermittencyof mmW multi-link communications

An open source LPWAN infrastructure called OpenChirpis discussed in [114] OpenChirp which is developed usingLoRWAN allows multiple users to provision and to managebattery-powered transducers across large areas like campusesindustrial zones or cities As pointed out in [30] [115]SDN plays a vital role in improving MEC type technologiesby removing the technical shortcomings in edge computingimplementations The authors summarize the work performedfor implementing MEC based on NFV and SDN where theSDN controller manages the communication between MECservers which form a data center at the edge Table VIIIsummarizes the reviewed state-of-the-art communication is-sues and solutions in MEC enabled IoT

12

TABLE VIII Comparison of the reviewed state-of-the-art communication issues and solutions in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Commnetworkarchitecture

Commresourceallocation

[109] Performance analysis of radio resource allocation in F-RANs for edge cache andadaptive model selection to improve spectral efficiency and energy efficiency

Low latency andhigh reliability

X

[112][113]

Use of mmWave spectrum for high data rate access to MEC servers and backhaul links Low latency andhigh reliability

X

[114] An open source LPWAN infrastructure which allows multiple users to provision andmanage battery-powered transducers across large areas

LPWANinfrastructure

X

[42] A virtualized edge computing architecture with a proxy VM migration scheme tominimize traffic in the core network

IoT big datastreams

X

[115] Proposed network architecture includes multi-interface wireless access network(eg FiWi) heterogeneous backhauling distributed cloudlets hierarchical structure ofa cloudlet and the SDN based mobile core network

IoT big datastreams

X

[110] A novel unified resource management scheme for Ethernet-based FiWi networks thatjointly allocates bandwidth for transmissions of both conventional broadband traffic andMEC data in a TDMA fashion

Mission-criticalIoT

X

[116] Introduce Mobile-IoT-Federation-as-a-Service (MIFaaS) to enable dynamic cooperationamong privatepublic local clouds of IoT devices at the edge of the cellular infrastructureThe selection of the best configuration of federated IoT cloud platforms are modeledas a coalition formation problem

Cellular IoT X

[117] Allocation of radio resources in a joint LTE and NB-IoT system based of MIFaasparadigm [116] Discovered that in handling high-end IoT data traffic a combinationbetween NB-IoT and LTE is essential in providing the needed high data rate and lowlatency

Mission-criticalIoT

X

[118] Integration of D2D communications into edge computing environment reduce transmis-sion delay and traffic load across the network

Mission-criticalIoT

X

[119] Use the theories of stochastic geometry queueing and parallel computing for provi-sioning and planning MEC networks

Communicationlatency

X

C Computation Offloading and Resource Allocation

1) Requirements Computation offloading is the mostprominent and widely discussed feature of MEC that empow-ers resource-constrained IoT devices with augmented com-putational capabilities [29] [33] This will not only prolongthe battery life of the IoT sensor nodes but also reduceE2E latency needed to run sophisticated applications In thefirst place UE has to decide whether to execute the rela-tively simple tasks locally or offload to the MEC servers(ie task model for binary offloading) [32] Secondly thedecision of computation offloading to the MEC servers can beperformed fully or partially In the partial offloading a subsetof computations is executed locally while the rest is offloadedto the MEC server by considering several factors such asusers or application preferences (eg application buffer state)radio and backhaul connections quality (ie between UEand MEC servers) UE capabilities or cloud capabilities andavailability [29]

The sole objective of the offloading policies need to be theminimization of execution delay Other critical concerns areto define the dependency of offloadable components of theapplications based on their ability to partition data (eg real-time user input has to be processed at UE without offloading)and to predict the execution time of multiple tasks Theexecution order or routines have to be carefully formulatedsince certain outcomes can be the inputs of other tasks Aspointed out in [32] the task models for partial offloading canbe represented by task-call graphs with sequential paralleland general dependencies

Although in MEC computation offloading enables power-ful cloud services at the edge level the insufficient batteryenergy at the tiny IoT devices may incur new challenges In

applications like IoT surveillance or remote asset managementthe nodes are typically hard to reach Those applications mayalso require to offload data more frequently in small chunksby consuming more energy Therefore it is necessary to con-sider not only the trade-off between energy consumption andexecution delay in both full and partial offloading scenarios inMEC but also the trade-off between computation energy andtransmission energy consumption in order to extend batterylife

The joint computation and communication resource alloca-tion should be properly addressed in order to get the maximumutilization of available resources Single MEC server will beallocated for the applications which cannot be partitionedThe resources in multiple MEC servers are allocated for theoffloaded applications that can be split into several parts Whena job arrives at the MEC server if there are enough resourcesthe scheduler has to allocate the VM for further processingIf there are no sufficient computation resources it delegatesthe task to the centralized cloud MEC servers also have toallocate computation and communication resources for userapplication jobs and MEC service jobs User mobility networktopology network scalability and load balancing are someother factors to be considered in order to define fare resourceutilization policies on MEC servers Specifically when IoTgateways share limited bandwidth among multiple IoT deviceswhich can handle video audio or bio-medical signals theallocation of bandwidth will become challenging [120] Thelow power wireless technologies (eg BLE ZigBee lowpower Wi-Fi and LPWAN standards like LoRA or SigFox)used in IoT networks have limited bandwidth When the IoTdevices access the MEC server which is acting as the IoTgateway they have to utilize either of those low-power wireless

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

4

TABLE III Summary of important acronyms

Acronym Definition Acronym Definition

3GPP Third Generation Partnership Project 5G Fifth Generation Wireless NetworkAI Artifical Intelligence AR Augmented RealityBLE Bluetooth Low Energy CaPC Cloud-aware Power ControlCPS Cyber Physical System C-RAN Cloud Radio Access NetworkD2D Device-to-device DDoS Distributed Denial of ServiceDoS Denial of Service E2E End-to-endEC Edge Computing eMBB enhance Mobile BroadbandEMM Energy-aware Mobility Management eNodeB Evolved Node BETSI European Telecommunications Standards Institute EU European UnionFiWi Fiber-enable Wireless F-RAN Fog Radio Access NetworkGDPR General Data Protection Regulation ICN Information Centric NetworkingICT Information Communication Technology IIoT Industrial Internet of ThingsIoT Internet of Things ISG Industry Specification GroupKDN Knowledge-Defined Networking LPWAN Low-power Wide Area NetworkLTE Long Term Evolution M2M Machine-to-machineMANO Management and Orchestration MCC Mobile Cloud ComputingMEC Multi-Access Edge Computing MIFaaS Mobile-IoT-Federation-as-a-ServiceMitM Man-in-the-Middle mmW millimeter-WaveMR Mixed Reality NB-IoT Narrow-band IoTNFV Network Function Virtualization PbD Privacy by DesignQoE Quality of Experience QoS Quality of ServiceRAN Radio Access Networks RAT Radio Access TechnologyRFID Radio-Frequency Identification RNC Radio Network ControllerSCeNB Small Cell eNodeBs SDLB Software Load BalancerSDN Software Defined Networking SDP Software Defined PrivacySIoT Social Internet of Things TDMA Time-division Multiple AccessUAV Unmanned Aerial Vehicles UE User EquipmentV2V Vehicle to Vehicle V2X Vehicle to EverythingVANET Vehicular Ad-hoc Network VM Virtual MachineVNF Virtual Network Function VR Virtual RealityVRARA Virtual RealityAugmented Reality Association WAN Wide Area NetworkingWAP Wireless Access Point WIoT Wearable Internet of ThingsWLAN Wirless Local Area Networking WSN Wireless Sensor Network

Finally Section VII concludes the paper We provide thedefinitions of frequently used acronyms in Table III

II IOT AND MEC APPLICATION SCENARIOS

This section focuses on how IoT can leverage MEC tech-nology in various application scenarios IoT itself is a classicapplication of MEC where the key value proposition of MECis exemplified in a variety of application scenarios (Figure 3)These values become evident in the utility factor measured bythe end user experience while using such IoT related services

Table IV and V respectively show the characteristics of dif-ferent IoT applications and how each application benefits fromMEC-IoT integration In addition Table VI summarizes thereviewed state-of-the-art applications in MEC-IoT domains

A Smart home and Smart city

One of the pioneering applications of the IoT technologyhas been in the areas of home automation and consumerelectronics [39] Several smart home applications that are builton the basis of IoT concept are already available in mostconsumer markets These range from the simple thermostatsensors to other more sophisticated automation systems likesmart metering smart heating and lighting cleaning servicesand home entertainment systems That notwithstanding theamount of data that would be generated on a typical IoTnetwork like the smart home is expected to be huge Hencetransferring such data to the centralized cloud servers will be

impractical with most pre-MEC techniques As a solutionMEC leverages specialized and reliable local services forprocessing and storage capabilities for the large IoT trafficcreated within a building The conventional gateways whichallow IoT applications to run on the centralized cloud can beempowered with MEC-server functionalities [40] [41] Thisextends gateway functionalities to the edge of the networkwith reduced communication latency Since such appliancesare statically deployed in smart home or smart building envi-ronments the cooperation with MEC servers will offer someother features such as easy instantiation relocation privacypreservation and upgrading when necessary [21] [42]

Correspondingly IoT technology has advanced from hometo community and even city scale applications We see nu-merous future promises for public safety health care utilitytourism and the transport sectors Enormous IoT data trafficproduced in smart cities can be ideally processed at the edgeof the network providing low latency and location aware-ness [43] [44] In particular a video cameras (ie deployedfor surveillance) connected with a Long Term Evolution (LTE)network can convey video streams to the MEC server forreal-time processing and anomaly detection [21] Collaborativeedge paradigms that connect multiple MEC servers (iededicated for different services) will advocate the applicationswhich need to process geographically distributed data Forinstance a connected health care application requires to col-laborate with entities from multiple domains such as hospitalpharmacy insurance logistics and government [45]

5

Fig 3 IoT and MEC application scenarios

TABLE IV Characteristics of Different IoT application

IoT Application Data type Data Capacity BackhaulConnectivity

Expected latency Number of IoT Devices

Smart home Stream Historical data

ge 10 MB of data per house-hold per day

Realtime 1 ms -1000 s ge10-100 per house

Smart city Stream Massive data

ge10-100 million GB of dataper city per day

Realtime le1ms ge1000-1million per city

Remote surgery [35] Stream data ge15 million per year Realtime le200 ms ge10-100 per surgeryRemote consultancy Stream data ge 500 million visits per year Realtime 1 ms-100 s 1-10 per appointmentAutonomous vehicles Stream

Massive datage 100 GB per vehicle per day Realtime le1 ms 50-200 per vehicle

AR [36] Stream Massive data

ge1 GBps Realtime le1 ms ge02 million globally

VR [36] Stream Massive data

ge1 GBps Realtime le1ms ge02 million globally

Gaming [36] Stream Massive data

ge10 Mbps Realtime le10 ms ge1 billion globally

Retail [37] Stream Historical data

100 Mbps - 1 Gbps RealtimeIntermittent

le1 ms ge100-1000 per shop

WIoT Stream data lt 1 GB per device Intermittent Several Hours ge1-10 per personFarming Historical data ge 1 GB per farm Intermittent Several hours 100-100000 per farmSmart energy Stream

Massive datage 100000 GB per day Realtime

Intermittent1ms - 10 mins ge 1 billion per grid

Industrial Internet [38] Stream Massive data

ge 100000 GB per day Realtime le1 ms ge 1 million per factory

6

TABLE V MEC and IoT benefits for each application

Required characteristicsof MEC and IoT

Description Smar

tho

me

Smar

tci

ty

Rem

ote

surg

ery

Rem

ote

heal

thco

nsul

tanc

y

Aut

onom

ous

vehi

cles

Aug

men

ted

Rea

lity

(AR

)

Vir

tual

Rea

lity

(VR

)

Gam

ing

Ret

ail

Wea

rabl

eIo

T

Farm

ing

Smar

ten

ergy

Indu

stri

alIn

tern

et

Low Latency Optimize to process a very high volume ofdata messages with minimal delay

X X X X X X X X X X

Increased Bandwidth Ability move a large set amount of datarapidly

X X X X X X X X X X X X

Content Awareness Adaptation of network characteristics ac-cording the local services requirements

X X X X X X X X X X

Low power devices Support for low power devices which haslimited transmission powers

X X X X X X X

Fixed wireless support Operation of wireless systems used to con-nect two fixed locations with a wireless link

X X X X X X X X X X

Fast inter-RAT handoff Speed up the handover takes place betweendifferent RATs

X X X X X X X X

Caching Keeping frequently accessed information ina location close to the requester

X X X X X X X

Edge Analytics An automated analytical computation is per-formed on data at a sensor network switchor other device instead of waiting for thedata to be sent back to a centralized datastore

X X X X X X X X X

Application virtualizationbetween edge and cloud

On demand application and service migra-tion from centralized cloud to the edgecloud

X X X X X X X X X X X

Private or local network Limit the communication and data ex-changes to a certain network segment

X X X X X X X X X X X X

Security Provide localized security X X X X X X XPrivacy Provide localized Privacy X X X X X X XFast Mobility Enable the ability to move or be moved fast

within the network or network coverablearea

X X X X X X X X

B Healthcare

Mobile health and telemedicine are identified as importantuse cases of 5G Wearable low power IoT medical sensorsfor monitoring health related data and tracking records arenow popular in public healthcare facilities [46] Although IoTtechnologies are widely adopted in the health sector [47]their performance goals will not be achievable without edgecomputing solutions like MEC [37] [48] [49] For instancehumanoid robots sitting next to an elderly person may needtactile feedback in 1ms latency for his or her care taking ser-vices Mission critical use cases like remote surgeries requireultra-low latency uninterrupted communication links andcollaborations among surgeons present in different locationsRemote patient monitoring is another use case which enablesconsultants in major cities to interact with patients residingfar away from the medical facility The frequent updates ofhealth records for an elderly person or someone with a chronicdisease needs to proceed ubiquitously and securely With suchpotential use cases and scenarios the role of MEC in healthand social assistance industries becomes more evident [37]

Some research works have already been published aboutthe cooperation between edge computing and IoT in thehealthcare sector In [50] authors describe a military health-

care service platform based on hierarchical IoT architectureand a semantic edge network model The hierarchical IoTarchitecture can collect the vital health parameters of thesoldiers their weapon status as well as their geographicallocations The control center of the battlefield performs therole of edge component which can process and store largeamount of health data sent over an SDN-based network Thepreliminary network architecture proposed in [51] providesreal-time context-aware collaboration for remote robotic tele-surgeries Big data analytics performed by edge computingare also important in e-Healthcare applications [52] In [53]Rahmani et al introduces the smart gateway concept foran IoT-based remote health monitoring system Here they exploit edge computing nodes to update the centralized cloudbased on the medical data generated by the IoT sensors Theirgeo-distributed network of smart e-Health gateways provideslocal data processing for real-time notification for medicalpractitioners secure and privacy preserved data gathering pa-tientsrsquo mobility network interoperability and energy efficientcommunication

C Autonomous VehiclesIoT Automotive5G is a key enabler of V2X (Vehicle to Everything) concept

which covers Vehicle to Vehicle (V2V) vehicle to infras-

7

tructure vehicle to device vehicle to pedestrian vehicle tohome and vehicle to grid [54] In the context of IoT Au-tomotive V2X requires critical communication infrastructurewhere reliability and ultra low latency are crucial factors [55]Use cases in these categories include autonomous and semi-autonomous driving vehicle maintenance and in vehicle info-tainment In order to operate an efficient and reliable vehicularnetwork several features have to be improved these includereal-time traffic monitoring [56] [57] continuous sensing invehicles [58] [59] support for Infotainment applications [60]and improved security [61] However these features cannot beserved by current mobile networks [62] In this vein upcoming5G mobile systems are expected to offer a higher level of flexi-bility leveraging the emerging technologies related to networksoftwarization [63] In this context V2X combined with MECprovides a viable and cost-effective solution that can acceleratedevelopment of V2X and IoT automotive systems [64]

It is important to improve the performance of RAN tech-nologies to enable IoT automatization MEC will play a vitalrole here also For instance MEC technologies may fulfillthe latency reliability and throughput requirements in V2Xchannel modeling of mmWave communication [65] Moreoverthe placement of the MEC server within the RAN providesflexible network services for the vehicle and to efficientlycontrol the radio network resources [66] It is also possibleto design a time-predicted handover mechanism for vehiclesby leveraging road side information at MEC server in order tomeet the demand for high mobility and reliability in vehicularnetworks [66]

In addition ICN-MEC integration can also tackle existingtechnical challenges such as massive mobility of vehicles scal-ability deployment strategies service orchestration massivedata handling fast big data processing as well as ensuringsecurity and privacy [67]

Unmanned aerial vehicles (UAVs) or drones are anothertype of autonomous vehicles which are capable of sensingits environment and navigating without human inputs UAVuse cases include but not limited to public safety smart agri-culture surveillance and environmental monitoring [68] Inorder to maximize the flight time the UAV battery life shouldbe essentially conserved by minimizing the overhead onboardWhen the required processing power exceeds the availableresources on UAV the application data can be offloaded toMEC Accompanying the advanced RATs MEC will facilitatethe offloading process from UAV due to its expected widedeployment in the network [68]

D Gaming AR and VR

Mixed reality (MR) combines virtual reality (VR) andaugmented reality (AR) technologies thereby enabling humansto interact more naturally with the virtual worlds based ondata aggregated by IoT devices [69] With IoT AR tech-nologies are able to benefit directly from the high end inter-connection of objects that characterizes the IoT environmentthrough which users can extend their interactions from thereal world to the virtual world [2] [70] Convergence ofVR and IoT can occur in many ways such as telepresence

tourism industry smart transportation networks and roboticassisted surgeries Exclusive AR and VR experiences withthe delivery of 360 navigable videos will be offered byenhanced mobile broadband connections with low latencyand high reliability for mission-critical services With present-day network standards this might be impossible to achievehowever with the predicted characteristics of 5G such as20 Gbps peak data rate and 1 ms round-trip over-the-airlatency this becomes more easily achievable As identified byETSI MEC will be an ideal solution for low-latency offloadservices in AR and VR applications that combine computergenerated data with physical reality [71] While operatingVR devices over wireless links and deploying the VR controlcenter at MEC server the tracking accuracy can be increasedwith round trip latency of 1 ms and high reliability [72]Migrating computationally intensive tasks to edge servers willincrease the computational capacity of VR devices and savetheir battery-life Furthermore MEC will allow VR devices toaccess cloud resources in an on-demand fashion [73]

MEC platforms provide high capacity and low latencywireless coverage for large venues like stadiums or smartcities with a massive density of users to enjoy the AR andVR experience For instance inside a smart building with anetwork of cameras obtaining raw video frames and preparingthe processed frames for display can be performed locallywith the help of edge computing Furthermore tracking thelocal position of the user or object building a model of theenvironment and identifying known objects in the environ-ment can be offloaded to the edge cloud Similarly in orderto get absolute experience of VR glasses the response timeshould be extremely low When the user moves his head hemay experience delay if the glasses need to access remotedata centers Therefore the expected interaction time betweenmachines and humans needs to be less than 1ms When thelatency of a VR application is more than 1ms the user willexperience cyber sickness which will be interrupting the realVR experience MEC servers in the nearest proximity will beable to serve such applications with ultra low latency Futuregames will be played beyond the entertainment purposes ontop of VR and AR applications which would require theminimum possible latency Pokmon Go and Ingress are twoexamples of successful games that combine AR and sensorinformation such as user location

E Retail

The second largest MEC use case is expected to be in the re-tail businesses [37] Currently IoT has dominated retail marketapplications in many ways including digital signage supplychain management intelligent payment solutions smart vend-ing machines shelves doors resource management stream-ing and safety The high class retail stores which use facialrecognition systems need high definition cameras that generatehuge volumes of data requiring powerful servers within thepremises Therefore the on-site MEC servers will assist toprocess these kind of large data sets produced by IoT devicesin a retail market Big data analytics in shopping centerscan further exploit the collaborative processing between edge

8

and cloud computing [52] Installation of MEC in a retailmarket also provides high speed mobile coverage throughoutthe store WiFi access points that are maintained per store canbe connected to the MEC server to provide WiFi connectivityfor store customers as needed The enabling of MEC willalso omit load balancing Wi-Fi controllers or policy enginesrequired in the wide area networks in the store Althoughnot many academic published research works are explicitlyfocusing on MEC and IoT [74] they have become enormouslyreputed and commercialized technologies in the industry andthe business sectors

F Wearable IoT (WIoT)

During the previous years wearable technology has evolvedtremendously from walkman to step trackers smart watchesto smart glasses The development of low power wirelesstechnologies such as BLE (Bluetooth Low Energy) fuels thedevelopment of wearable devices Present-day wearables spanfrom low-end devices such as health and fitness trackers tohigh-end devices such as VRAR helmets and smart watchesIt is expected that wearables will become the worlds best-selling consumer electronics product after smartphones witha global availability of more than 929 million devices by2021 [75] With the new application domains and enablingservices wearable devices will demand more sophisticatedcommunication infrastructures For instance VRAR wear-ables are demanding gigabits throughput network connectivityto run their applications On the other hand dense deploymentof wearable devices in smart cities will increase the networktraffic on communication networks Thus the next generationcommunication networks should be able to provide the gigabitexperience for the anticipated ultra dense wearable devices[76]

Although cloud computing has enabled wide range of newnetworking services it cannot alone fulfill the upcoming re-quirements for the future wearable ecosystem Mainly the cen-tralized cloud data centers fails due to long End-to-End (E2E)latency Delay-sensitive wearable applications such as VRperceptual stability requires ultra low delay In this contextMEC has the potential to solve the limitations in current cloudbased systems by combining cloud and MEC infrastructuresThis will enable providers deploy storage computing andcaching capabilities in close proximity with such wearabledevices [76]

G IoT in Mechanized Agriculture

In order to meet the demands for future food productionthe agricultural sector will require some major evolution whereIoT will be integrated in various production management andanalytical processes [77] [78] The present-day agriculturalsector has been slow to adopting the emerging Machine-to-Machine (M2M) and IoT technologies when compared withother sectors like smart cities and the medical fields [79]

Precision farming and smart agriculture can be achievedusing autonomous vehicles (tractors) remote monitoring andreal-time analytics It is reported that farmers are increasinglyturning to agricultural drones and satellites to survey their

lands and generate crop data IoT sensors may provide in-formation about crop yields rainfall pest infestation and soilnutrition which are invaluable to production and can improvefarming techniques over time Although low latency is not acritical requirement in smart farming environment manage-ment of large data sets will be a key requirement to considerMEC servers located on-site can assist high tech farmingby collecting and analyzing big data on agriculture in orderto maximize efficiency Likewise without moving everydayfarming applications to a remote cloud MEC platforms canbenefit in terms of data access synchronization storage andother overhead costs the farmer might normally incur

The use of IoT-based automated data collection and moni-toring systems in poultry houses can be used to increase workefficiency and service quality and get a deeper understandingof chicken nurturing [80] Sensing technologies can be usedin carbon dioxide and luminosity sensing these are importantparameters in large scale poultry houses Gas sensors canbe used to get all necessary information to prevent chickeninfertility due to problems such as low carbon dioxide levelsLuminosity senors can help to maintain the proper luminositylevel for optimum productivity Similar to smart farms lowlatency is not a critical requirement in smart poultry houses[80] However it is critical to manage large data sets whereon-site MEC servers can be used In addition sharing the databetween poultry houses and storing legacy data in centralizedservers are important in identifying abnormal incidents in thefarm [81] With the use of MEC poultry houses can workwith intermittent connectivity to the centralized clouds In thatcase MEC servers can temporarily hold the data until farmsare connected with the centralized clouds

H Smart Energy

The smart grid system is an Information CommunicationTechnology (ICT)-enabled energy generation transmissionand distribution network It has capabilities to continuouslysense analyze and monitor both energy flow and energytransportation infrastructure Such features are enabled byadding digital controls and enabling network monitoring andtelecommunication capabilities As a result a smart grid doesnot only provide two-way flows of electrical power but alsoenables real-time automated bidirectional flow of informa-tion Adding such smartness to the aging energy infrastructurewill foster a more efficient energy system

IoT is considered as the foundation for realizing intelli-gence capabilities in smart grid systems IoT integrates theInternet-connectivity into all kinds of grid components suchas transformers breakers switches meters relays intelligentelectronic devices capacitor banks voltage regulators cam-eras and many more These IoT devices are then used tocapture the data required to enable automations IoT-enabledsmart grids provide several benefits such as reduced capitalexpenditure optimized renewable capacity lowered mainte-nance costs and enhanced customer engagement On one handthe transformation of an electrical grid into a smart systemrequires nearly every device and piece of equipment to havebuilt-in secure interconnected intelligence On the other hand

9

an efficient system is required to manage the generated dataie transferring storing and analyzing such huge amounts ofdata which are collected from these smart devices Thereforecloud computing is a viable solution to these IoT-based smartgrids [90]

Generally smart grids are spanning over large geograph-ical areas They often confront bandwidth bottlenecks andcommunication delays due to poor network connectivity andvast number of devices generating data Thus the traditionalcentralized cloud architecture is not suitable for the domain ofthe smart grid since it relies heavily on centralized processing[91] Many delay sensitive smart grid applications such asfault detection isolation and service restoration or VoltVARoptimization cannot tolerate round trip delay to access cen-tralized cloud systems MEC is identified as the viable cloud

computing option to address these limitations MEC allowsthe computation to be performed closer to the data sourceMoreover the potential attack points for the grid is increasingwith the growth of ubiquitous sensor deployment Everysmart IoT device can be vulnerable to potential attacks MECprovides the opportunity to enforce security mechanism closerto the end devices As such even if an attacker gains accessto an endpoint device the attack gets no further informationbeyond the local network segment since MEC has capabilitiesto notice the intrusion and cease the accessibility [85]

I Industrial Internet

The Industrial Internet of Things (IIoT) also known asIndustry 40 [92] is an application of IoT in the domain ofmanufacturing IIoT incorporates numerous advanced commu-

TABLE VI The reviewed state-of-the-art MEC integration in different IoT applications

Ref Description Smar

tH

ome

Smar

tC

ity

Hea

lthca

re

IoT

Aut

omot

ive

Gam

ing

AR

VR

Ret

ail

Wea

rabl

eIo

T

Smar

tA

gric

ultu

re

Smar

tE

nerg

y

Indu

stri

alIn

tern

et

[40] Preliminary design of deploying MEC server functionalities in a smart hometo realize IoT gateway with direct M2M interaction in LTE networks

X

[41] Introduce Gateway-as-a-Service for heterogeneous IoT devices on top of thevirtualization technologies in edge computing

X X

[44] Propose an autonomic creation of MEC services to enhance QoS of videostreaming in smart cities

X

[50] Propose a semantic edge-based IoT architecture for military health services inbattlefield

X

[51] Provide a conceptual MEC based architecture for mission-critical context awarecollaboration in remote surgeries

X

[53] Describe and implement a smart e-Health gateway at the edge of the networksuitable for ubiquitous healthcare systems

X

[64] Analysis on research and engineering challenges co-existence of cloud edgecomputing and data caching strategies at the edge for vehicular networks

X

[82] Discuss the design aspects for the radio access in 5G V2X X[65] Discuss the benefits of merging MEC and mmWave technologies for 5G

applicationsX X X

[66] Propose a novel MEC-based architecture for future cellular vehicular networks X[67] Discuss the benefits of combining ICN and MEC in the context of connected

vehicle environmentsX

[52] Propose a framework for big data analytics between edge and cloud computingplatforms

X X

[74] Design and implement a fog computing based framework that support sharingand reusing contextual data across services in smart city and retail stores

X X

[83] Present a usecase of MEC for Tactile Internet based 5G gaming application X[84] A demonstration of MEC for Tactile Internet based 5G gaming application X[76] Discuss the role of MEC in 5G WIoT communication and its challenges X X[68] Propose an UAV-based IoT platform for a crowd surveillance use case X X[78] Develop and test a ubiquitous sensor network platform for crop lands automa-

tion maintenance in precision agricultureX

[70] Present a serverless edge computing architecture that enables the offloading ofmobile computation with low latency and high throughput using a mobile ARapplication

X

[85] Discuss the benefits of MEC and edge computing (EC) to enhance the securityof smart grids

X

[86] Present a method to optimize the EC based video streaming schemes forIndustrial IoT

X

[87] Present the use of edge computing to provide elastic resources and services toenable microdatabases architecture for IIoT

X

[88] Propose a fog-based communication architecture for Industry 40 applications X[73] Describe research directions and enablers of wireless interconnected VR

systemsX

[89] Design an optimization framework for VRAR communication via small-cellcooperation

X

10

nication and automation technologies such as M2M commu-nication machine learning and big data analytics to improveintelligence and the connectivity [93] For instance IIoTnetworks can connect all of the employees data and processesfrom the factory floor and forward them to the executiveoffices Thus decision makers or employees can create a fulland accurate view of their manufacturing process by usingIIoT network hence improving their ability to make moreinformed decisions IIoT also helps the exploitation as well asimplementation of new intelligent technologies to acceleratethe innovation and transformation of the factory workforce[92]

Primarily IIoT is seen as a way to improve operationalefficiency However IIoT provides a wide range of otherbenefits such as improving connectivity efficiency scalabilitytime savings as well as cost savings for manufacturing pro-cesses with the maximum use of smart machines [92] [94] Ingeneral these smart machines operate with higher accuracygreater efficiency and constant working capabilities than hu-mans [95] Thus IIoT has great potential for improving qualitycontrol sustainability and overall supply chain efficiency

MEC will play a vital role in enabling future IIoT applica-tions [96] by addressing the shortcomings of M2M communi-cation (eg latency resilience cost peer-to-peer connectivitysecurity) in IIoT domain [97] [98] Current market trendsalready show that edge computing will represent many im-plementation scenarios for IIoT For instance real-time edgeanalytics and enhanced edge security are two key drivers inthe creation of new IIoT deployments Thus the addition ofMEC in IIoT networks will fuel the evolution of IIoT as wellas create new business applications [99]

One way to optimize the use of conventional edge com-puting in video streaming schemes for IIoT is presented in[86] By using machine learning algorithms edge computingcan process the sensor data before transmitting to the cloudThis mitigates against the degradation of service quality of thevideo streaming Aggregation of all the sensor data to a singledata center increases latency and raises performance concernsin IIoT domain In order to solve this issue a microdatabasearchitecture is proposed for the Industrial Internet [87] It holdsthe data close to the industrial processes but also makes itavailable near the applications that can benefit from the dataEdge computing also provides elastic resources and services toenable micro-database architecture [87] A fog-based commu-nication architecture for Industry 40 applications is proposedin [88] This approach will substantially minimize the energyconsumption of the IoT nodes Edge computational capabilitiesare further used to predict future data measurements andreduce the throughput from IoT devices to the control unit

III TECHNICAL ASPECTS OF MEC ENABLED IOT

To realize the MEC exploitation for IoT applications thekey value propositions are mostly seen from the technicalparameters such as scalability communication computationoffloading and resource allocation mobility management se-curity privacy and trust management This section describesthe state-of-the-art of each of these technical parameters hence

giving a clear background against which the benefits of MECcan be envisioned

A Scalability

1) Requirements When it comes to actual deployment ofMEC platform for IoT systems scalability is a key factorto consider The compatibility of MEC servers to multiplenetwork environments is one of the factors that will driveits large scale adoption in future networks [100] The IoTenvironment will consist of hundreds of billions of sensors ac-tuators Radio-Frequency Identification (RFID)-tagged objectssoftware vehicles and embedded systems all interconnectedin a huge network of cyber-physical systems At a utility scaleconsideration these devices will be working in close collab-oration to deliver the expected services in technologies likethe smart grids virtual power plants smart homes intelligenttransportation and smart cities That being said the role ofscalability to the realization of such a hyper-connected IoTenvironment becomes more obvious The IoT environment willrequire a dynamic range of capabilities in the network space ifsuch large numbers of devices are to be supported effectively

2) Related work Currently MEC servers have been con-firmed to be compatible with LTE macro base station (eN-odeB) sites 3G Radio Network Controller (RNC) site multi-Radio Access Technology (RAT) cell aggregation site andat the edge of the core network [2] Such multi-RAT cellaggregation schemes can be implemented indoor or outdoorsettings depending on the requirements This invariably en-ables MEC to be applied to many different possible scenariosThe larger the deployment scenarios for MEC the more therange of capabilities it can handle this also translates to higherscalability for MEC-enable technologies like IoT

Designing an edge cloud network implies that an optimallocation for citing the cloud facility is first determined In[105] authors present a design optimization scheme for theMEC architecture based on link-path formulation supportedby heuristics in order to optimize the computation time forthe scheme In this approach consideration is given to bothusers and VMs mobility Hence an optimal point to installthe MEC server is determined through a tread-off betweeninstallation cost and the quality of service to be deliveredTable VII compares the reviewed state-of-the-art scalabilityfeature in MEC enabled IoT

B Communication

1) Requirements There are three main categories for thecommunication concerns about MEC [106] Wireless accesswhile offloading to the mobile edge host Backhaul accesswhile offloading to a remote cloud server Communicationamong IoT devices mobile edge host and remote cloudservers when they collaboratively execute multiple jobs Thefirst and the second categories are the most renowned on behalfof the MEC servers which are the small scale data centersdeployed by the network operators and can be co-located withthe Wireless Access Points (WAPs) In the IoT supportiveMEC systems the consumer devices may communicate with

11

TABLE VII Comparison of the reviewed state-of-the-art scalability feature in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Addressing Search

[101] Discusses the challenges in searching imposed by the burgeoning field of IoT General IoT X[94] Examines a variety of popular and innovative IoT solutions in terms of context-aware

technology perspectives to serve as a conceptual framework for context-aware productdevelopment and research in the IoT paradigm

General IoT solu-tions

X

[102] Proposes an innovative distributed architecture combining machine-to-machine industry-mature protocols (ie MQTT and CoAP) in an original way to enhance the scalabilityof gateways for the efficient IoT-cloud integration

IoT cloud inte-gration

X X

[103] Studies an implementation of edge computing which exploits transparent computing tobuild scalable IoT platforms using transparent computing

Wearable IoT X

[104] Introduces a lightweight edge gateway for the IoT architecture using container-basedvirtualization techniques

General IoT X

the MEC servers either directly or with the support of neigh-boring devices using Device-to-Device (D2D) communicationFor the third category WAPs enable access to the remote datacenters in the central cloud through backhaul links

In order to reap the maximum advantage of computationoffloading leveraged at the edge servers MEC systems needefficient communication channels Unlike the wired connec-tions in the conventional grid computing and cloud computingthe wireless access links between the mobile devices andcloud computing resources in the edge computing paradigmcan be unstable Sudden service outages may occur with theinterruption of access links The inherent challenges withwireless communication channels like multi-path fading in-terference and spectrum shortage should always be taken intoaccount for the design of MEC systems to seamlessly integratecomputation offloading and radio resource management [32]Moreover both wireless and backhaul access links have lim-ited capacities which should be properly shared among mobiledevices in a similar way as sharing the computing resources ofthe MEC server Hence having a cooperative scheme for thejoint allocation of communication and computation resourcesis important for the successful deployment of MEC [106]Redesigning both communication and networking protocolsto integrate communication infrastructures in MEC and IoTsystems is a challenging task The key focus should be onimproving the computation efficiency with respect to datatransmission

Another major requirement is to maintain interoperabilitywhile addressing heterogeneous communication technologiesthat have to be utilized in IoT and MEC paradigms in 5GThere are plenty of radio technologies that facilitate IoT Low-Power Wide Area Networks (LPWANs) (eg WCDMA LTEnarrowband IoT (NB-IoT) Wi-Fi Bluetooth Zigbee SIGFOXand LoRA) The choice of these LPWAN technologies maycreate trade-offs among signal strength operational rangethroughput and power consumption With the arrival of 5Gthe convergence of these communication technologies needsto be achieved since one network will not be fitting based onthose trade-offs

2) Related work Recently Fog-Radio Access Network (F-RAN) was introduced by Peng et al to consolidate the hetero-geneous networks into a single network architecture with 5Geven though they do not operate in the same bands to gain highspectral and operating and energy efficiency [107] Well known

Cloud Radio Access Network (C-RAN) architecture can per-form cooperative transmission across multiple edge nodes withcentralized cloud computing servers via fronthaul links [108]Although C-RAN provides high spectral efficiencies due tothe enhanced interference management capabilities with thecentralized baseband processing at the cloud it has potentiallylarge latencies F-RAN is proposed for 5G MEC deploymentsas an advanced socially aware mobile networking architectureto provide high spectral efficiency while maintaining highenergy efficiency and low latency [107] [108] Precodingdesign resource block allocation user scheduling and cellassociation are jointly designed for radio resource allocation inF-RANs in order to optimize spectral and energy efficienciesand latency performances [109] In [110] Rimal et al pro-pose a unified Time-Division Multiple Access (TDMA) basedresource management scheme for offloading traffic over Fiber-enabled Wireless (FiWi) access networks

In the envisioned 5G systems and MEC architecture bothbackhaul and wireless access links can be facilitated bymillimeter-Wave (mmW) spectrum [111] The use of mmWspectrum will enable high data rate access to MEC function-alities with low latency On the other hand MEC provideslocal computation power usefully for optimizing the perfor-mance of mmW communications In [112] [113] the authorsaddress the joint optimization of communicationcomputationresources with mmW communication They have taken the ad-vantage of blocking probabilities by considering intermittencyof mmW multi-link communications

An open source LPWAN infrastructure called OpenChirpis discussed in [114] OpenChirp which is developed usingLoRWAN allows multiple users to provision and to managebattery-powered transducers across large areas like campusesindustrial zones or cities As pointed out in [30] [115]SDN plays a vital role in improving MEC type technologiesby removing the technical shortcomings in edge computingimplementations The authors summarize the work performedfor implementing MEC based on NFV and SDN where theSDN controller manages the communication between MECservers which form a data center at the edge Table VIIIsummarizes the reviewed state-of-the-art communication is-sues and solutions in MEC enabled IoT

12

TABLE VIII Comparison of the reviewed state-of-the-art communication issues and solutions in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Commnetworkarchitecture

Commresourceallocation

[109] Performance analysis of radio resource allocation in F-RANs for edge cache andadaptive model selection to improve spectral efficiency and energy efficiency

Low latency andhigh reliability

X

[112][113]

Use of mmWave spectrum for high data rate access to MEC servers and backhaul links Low latency andhigh reliability

X

[114] An open source LPWAN infrastructure which allows multiple users to provision andmanage battery-powered transducers across large areas

LPWANinfrastructure

X

[42] A virtualized edge computing architecture with a proxy VM migration scheme tominimize traffic in the core network

IoT big datastreams

X

[115] Proposed network architecture includes multi-interface wireless access network(eg FiWi) heterogeneous backhauling distributed cloudlets hierarchical structure ofa cloudlet and the SDN based mobile core network

IoT big datastreams

X

[110] A novel unified resource management scheme for Ethernet-based FiWi networks thatjointly allocates bandwidth for transmissions of both conventional broadband traffic andMEC data in a TDMA fashion

Mission-criticalIoT

X

[116] Introduce Mobile-IoT-Federation-as-a-Service (MIFaaS) to enable dynamic cooperationamong privatepublic local clouds of IoT devices at the edge of the cellular infrastructureThe selection of the best configuration of federated IoT cloud platforms are modeledas a coalition formation problem

Cellular IoT X

[117] Allocation of radio resources in a joint LTE and NB-IoT system based of MIFaasparadigm [116] Discovered that in handling high-end IoT data traffic a combinationbetween NB-IoT and LTE is essential in providing the needed high data rate and lowlatency

Mission-criticalIoT

X

[118] Integration of D2D communications into edge computing environment reduce transmis-sion delay and traffic load across the network

Mission-criticalIoT

X

[119] Use the theories of stochastic geometry queueing and parallel computing for provi-sioning and planning MEC networks

Communicationlatency

X

C Computation Offloading and Resource Allocation

1) Requirements Computation offloading is the mostprominent and widely discussed feature of MEC that empow-ers resource-constrained IoT devices with augmented com-putational capabilities [29] [33] This will not only prolongthe battery life of the IoT sensor nodes but also reduceE2E latency needed to run sophisticated applications In thefirst place UE has to decide whether to execute the rela-tively simple tasks locally or offload to the MEC servers(ie task model for binary offloading) [32] Secondly thedecision of computation offloading to the MEC servers can beperformed fully or partially In the partial offloading a subsetof computations is executed locally while the rest is offloadedto the MEC server by considering several factors such asusers or application preferences (eg application buffer state)radio and backhaul connections quality (ie between UEand MEC servers) UE capabilities or cloud capabilities andavailability [29]

The sole objective of the offloading policies need to be theminimization of execution delay Other critical concerns areto define the dependency of offloadable components of theapplications based on their ability to partition data (eg real-time user input has to be processed at UE without offloading)and to predict the execution time of multiple tasks Theexecution order or routines have to be carefully formulatedsince certain outcomes can be the inputs of other tasks Aspointed out in [32] the task models for partial offloading canbe represented by task-call graphs with sequential paralleland general dependencies

Although in MEC computation offloading enables power-ful cloud services at the edge level the insufficient batteryenergy at the tiny IoT devices may incur new challenges In

applications like IoT surveillance or remote asset managementthe nodes are typically hard to reach Those applications mayalso require to offload data more frequently in small chunksby consuming more energy Therefore it is necessary to con-sider not only the trade-off between energy consumption andexecution delay in both full and partial offloading scenarios inMEC but also the trade-off between computation energy andtransmission energy consumption in order to extend batterylife

The joint computation and communication resource alloca-tion should be properly addressed in order to get the maximumutilization of available resources Single MEC server will beallocated for the applications which cannot be partitionedThe resources in multiple MEC servers are allocated for theoffloaded applications that can be split into several parts Whena job arrives at the MEC server if there are enough resourcesthe scheduler has to allocate the VM for further processingIf there are no sufficient computation resources it delegatesthe task to the centralized cloud MEC servers also have toallocate computation and communication resources for userapplication jobs and MEC service jobs User mobility networktopology network scalability and load balancing are someother factors to be considered in order to define fare resourceutilization policies on MEC servers Specifically when IoTgateways share limited bandwidth among multiple IoT deviceswhich can handle video audio or bio-medical signals theallocation of bandwidth will become challenging [120] Thelow power wireless technologies (eg BLE ZigBee lowpower Wi-Fi and LPWAN standards like LoRA or SigFox)used in IoT networks have limited bandwidth When the IoTdevices access the MEC server which is acting as the IoTgateway they have to utilize either of those low-power wireless

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

5

Fig 3 IoT and MEC application scenarios

TABLE IV Characteristics of Different IoT application

IoT Application Data type Data Capacity BackhaulConnectivity

Expected latency Number of IoT Devices

Smart home Stream Historical data

ge 10 MB of data per house-hold per day

Realtime 1 ms -1000 s ge10-100 per house

Smart city Stream Massive data

ge10-100 million GB of dataper city per day

Realtime le1ms ge1000-1million per city

Remote surgery [35] Stream data ge15 million per year Realtime le200 ms ge10-100 per surgeryRemote consultancy Stream data ge 500 million visits per year Realtime 1 ms-100 s 1-10 per appointmentAutonomous vehicles Stream

Massive datage 100 GB per vehicle per day Realtime le1 ms 50-200 per vehicle

AR [36] Stream Massive data

ge1 GBps Realtime le1 ms ge02 million globally

VR [36] Stream Massive data

ge1 GBps Realtime le1ms ge02 million globally

Gaming [36] Stream Massive data

ge10 Mbps Realtime le10 ms ge1 billion globally

Retail [37] Stream Historical data

100 Mbps - 1 Gbps RealtimeIntermittent

le1 ms ge100-1000 per shop

WIoT Stream data lt 1 GB per device Intermittent Several Hours ge1-10 per personFarming Historical data ge 1 GB per farm Intermittent Several hours 100-100000 per farmSmart energy Stream

Massive datage 100000 GB per day Realtime

Intermittent1ms - 10 mins ge 1 billion per grid

Industrial Internet [38] Stream Massive data

ge 100000 GB per day Realtime le1 ms ge 1 million per factory

6

TABLE V MEC and IoT benefits for each application

Required characteristicsof MEC and IoT

Description Smar

tho

me

Smar

tci

ty

Rem

ote

surg

ery

Rem

ote

heal

thco

nsul

tanc

y

Aut

onom

ous

vehi

cles

Aug

men

ted

Rea

lity

(AR

)

Vir

tual

Rea

lity

(VR

)

Gam

ing

Ret

ail

Wea

rabl

eIo

T

Farm

ing

Smar

ten

ergy

Indu

stri

alIn

tern

et

Low Latency Optimize to process a very high volume ofdata messages with minimal delay

X X X X X X X X X X

Increased Bandwidth Ability move a large set amount of datarapidly

X X X X X X X X X X X X

Content Awareness Adaptation of network characteristics ac-cording the local services requirements

X X X X X X X X X X

Low power devices Support for low power devices which haslimited transmission powers

X X X X X X X

Fixed wireless support Operation of wireless systems used to con-nect two fixed locations with a wireless link

X X X X X X X X X X

Fast inter-RAT handoff Speed up the handover takes place betweendifferent RATs

X X X X X X X X

Caching Keeping frequently accessed information ina location close to the requester

X X X X X X X

Edge Analytics An automated analytical computation is per-formed on data at a sensor network switchor other device instead of waiting for thedata to be sent back to a centralized datastore

X X X X X X X X X

Application virtualizationbetween edge and cloud

On demand application and service migra-tion from centralized cloud to the edgecloud

X X X X X X X X X X X

Private or local network Limit the communication and data ex-changes to a certain network segment

X X X X X X X X X X X X

Security Provide localized security X X X X X X XPrivacy Provide localized Privacy X X X X X X XFast Mobility Enable the ability to move or be moved fast

within the network or network coverablearea

X X X X X X X X

B Healthcare

Mobile health and telemedicine are identified as importantuse cases of 5G Wearable low power IoT medical sensorsfor monitoring health related data and tracking records arenow popular in public healthcare facilities [46] Although IoTtechnologies are widely adopted in the health sector [47]their performance goals will not be achievable without edgecomputing solutions like MEC [37] [48] [49] For instancehumanoid robots sitting next to an elderly person may needtactile feedback in 1ms latency for his or her care taking ser-vices Mission critical use cases like remote surgeries requireultra-low latency uninterrupted communication links andcollaborations among surgeons present in different locationsRemote patient monitoring is another use case which enablesconsultants in major cities to interact with patients residingfar away from the medical facility The frequent updates ofhealth records for an elderly person or someone with a chronicdisease needs to proceed ubiquitously and securely With suchpotential use cases and scenarios the role of MEC in healthand social assistance industries becomes more evident [37]

Some research works have already been published aboutthe cooperation between edge computing and IoT in thehealthcare sector In [50] authors describe a military health-

care service platform based on hierarchical IoT architectureand a semantic edge network model The hierarchical IoTarchitecture can collect the vital health parameters of thesoldiers their weapon status as well as their geographicallocations The control center of the battlefield performs therole of edge component which can process and store largeamount of health data sent over an SDN-based network Thepreliminary network architecture proposed in [51] providesreal-time context-aware collaboration for remote robotic tele-surgeries Big data analytics performed by edge computingare also important in e-Healthcare applications [52] In [53]Rahmani et al introduces the smart gateway concept foran IoT-based remote health monitoring system Here they exploit edge computing nodes to update the centralized cloudbased on the medical data generated by the IoT sensors Theirgeo-distributed network of smart e-Health gateways provideslocal data processing for real-time notification for medicalpractitioners secure and privacy preserved data gathering pa-tientsrsquo mobility network interoperability and energy efficientcommunication

C Autonomous VehiclesIoT Automotive5G is a key enabler of V2X (Vehicle to Everything) concept

which covers Vehicle to Vehicle (V2V) vehicle to infras-

7

tructure vehicle to device vehicle to pedestrian vehicle tohome and vehicle to grid [54] In the context of IoT Au-tomotive V2X requires critical communication infrastructurewhere reliability and ultra low latency are crucial factors [55]Use cases in these categories include autonomous and semi-autonomous driving vehicle maintenance and in vehicle info-tainment In order to operate an efficient and reliable vehicularnetwork several features have to be improved these includereal-time traffic monitoring [56] [57] continuous sensing invehicles [58] [59] support for Infotainment applications [60]and improved security [61] However these features cannot beserved by current mobile networks [62] In this vein upcoming5G mobile systems are expected to offer a higher level of flexi-bility leveraging the emerging technologies related to networksoftwarization [63] In this context V2X combined with MECprovides a viable and cost-effective solution that can acceleratedevelopment of V2X and IoT automotive systems [64]

It is important to improve the performance of RAN tech-nologies to enable IoT automatization MEC will play a vitalrole here also For instance MEC technologies may fulfillthe latency reliability and throughput requirements in V2Xchannel modeling of mmWave communication [65] Moreoverthe placement of the MEC server within the RAN providesflexible network services for the vehicle and to efficientlycontrol the radio network resources [66] It is also possibleto design a time-predicted handover mechanism for vehiclesby leveraging road side information at MEC server in order tomeet the demand for high mobility and reliability in vehicularnetworks [66]

In addition ICN-MEC integration can also tackle existingtechnical challenges such as massive mobility of vehicles scal-ability deployment strategies service orchestration massivedata handling fast big data processing as well as ensuringsecurity and privacy [67]

Unmanned aerial vehicles (UAVs) or drones are anothertype of autonomous vehicles which are capable of sensingits environment and navigating without human inputs UAVuse cases include but not limited to public safety smart agri-culture surveillance and environmental monitoring [68] Inorder to maximize the flight time the UAV battery life shouldbe essentially conserved by minimizing the overhead onboardWhen the required processing power exceeds the availableresources on UAV the application data can be offloaded toMEC Accompanying the advanced RATs MEC will facilitatethe offloading process from UAV due to its expected widedeployment in the network [68]

D Gaming AR and VR

Mixed reality (MR) combines virtual reality (VR) andaugmented reality (AR) technologies thereby enabling humansto interact more naturally with the virtual worlds based ondata aggregated by IoT devices [69] With IoT AR tech-nologies are able to benefit directly from the high end inter-connection of objects that characterizes the IoT environmentthrough which users can extend their interactions from thereal world to the virtual world [2] [70] Convergence ofVR and IoT can occur in many ways such as telepresence

tourism industry smart transportation networks and roboticassisted surgeries Exclusive AR and VR experiences withthe delivery of 360 navigable videos will be offered byenhanced mobile broadband connections with low latencyand high reliability for mission-critical services With present-day network standards this might be impossible to achievehowever with the predicted characteristics of 5G such as20 Gbps peak data rate and 1 ms round-trip over-the-airlatency this becomes more easily achievable As identified byETSI MEC will be an ideal solution for low-latency offloadservices in AR and VR applications that combine computergenerated data with physical reality [71] While operatingVR devices over wireless links and deploying the VR controlcenter at MEC server the tracking accuracy can be increasedwith round trip latency of 1 ms and high reliability [72]Migrating computationally intensive tasks to edge servers willincrease the computational capacity of VR devices and savetheir battery-life Furthermore MEC will allow VR devices toaccess cloud resources in an on-demand fashion [73]

MEC platforms provide high capacity and low latencywireless coverage for large venues like stadiums or smartcities with a massive density of users to enjoy the AR andVR experience For instance inside a smart building with anetwork of cameras obtaining raw video frames and preparingthe processed frames for display can be performed locallywith the help of edge computing Furthermore tracking thelocal position of the user or object building a model of theenvironment and identifying known objects in the environ-ment can be offloaded to the edge cloud Similarly in orderto get absolute experience of VR glasses the response timeshould be extremely low When the user moves his head hemay experience delay if the glasses need to access remotedata centers Therefore the expected interaction time betweenmachines and humans needs to be less than 1ms When thelatency of a VR application is more than 1ms the user willexperience cyber sickness which will be interrupting the realVR experience MEC servers in the nearest proximity will beable to serve such applications with ultra low latency Futuregames will be played beyond the entertainment purposes ontop of VR and AR applications which would require theminimum possible latency Pokmon Go and Ingress are twoexamples of successful games that combine AR and sensorinformation such as user location

E Retail

The second largest MEC use case is expected to be in the re-tail businesses [37] Currently IoT has dominated retail marketapplications in many ways including digital signage supplychain management intelligent payment solutions smart vend-ing machines shelves doors resource management stream-ing and safety The high class retail stores which use facialrecognition systems need high definition cameras that generatehuge volumes of data requiring powerful servers within thepremises Therefore the on-site MEC servers will assist toprocess these kind of large data sets produced by IoT devicesin a retail market Big data analytics in shopping centerscan further exploit the collaborative processing between edge

8

and cloud computing [52] Installation of MEC in a retailmarket also provides high speed mobile coverage throughoutthe store WiFi access points that are maintained per store canbe connected to the MEC server to provide WiFi connectivityfor store customers as needed The enabling of MEC willalso omit load balancing Wi-Fi controllers or policy enginesrequired in the wide area networks in the store Althoughnot many academic published research works are explicitlyfocusing on MEC and IoT [74] they have become enormouslyreputed and commercialized technologies in the industry andthe business sectors

F Wearable IoT (WIoT)

During the previous years wearable technology has evolvedtremendously from walkman to step trackers smart watchesto smart glasses The development of low power wirelesstechnologies such as BLE (Bluetooth Low Energy) fuels thedevelopment of wearable devices Present-day wearables spanfrom low-end devices such as health and fitness trackers tohigh-end devices such as VRAR helmets and smart watchesIt is expected that wearables will become the worlds best-selling consumer electronics product after smartphones witha global availability of more than 929 million devices by2021 [75] With the new application domains and enablingservices wearable devices will demand more sophisticatedcommunication infrastructures For instance VRAR wear-ables are demanding gigabits throughput network connectivityto run their applications On the other hand dense deploymentof wearable devices in smart cities will increase the networktraffic on communication networks Thus the next generationcommunication networks should be able to provide the gigabitexperience for the anticipated ultra dense wearable devices[76]

Although cloud computing has enabled wide range of newnetworking services it cannot alone fulfill the upcoming re-quirements for the future wearable ecosystem Mainly the cen-tralized cloud data centers fails due to long End-to-End (E2E)latency Delay-sensitive wearable applications such as VRperceptual stability requires ultra low delay In this contextMEC has the potential to solve the limitations in current cloudbased systems by combining cloud and MEC infrastructuresThis will enable providers deploy storage computing andcaching capabilities in close proximity with such wearabledevices [76]

G IoT in Mechanized Agriculture

In order to meet the demands for future food productionthe agricultural sector will require some major evolution whereIoT will be integrated in various production management andanalytical processes [77] [78] The present-day agriculturalsector has been slow to adopting the emerging Machine-to-Machine (M2M) and IoT technologies when compared withother sectors like smart cities and the medical fields [79]

Precision farming and smart agriculture can be achievedusing autonomous vehicles (tractors) remote monitoring andreal-time analytics It is reported that farmers are increasinglyturning to agricultural drones and satellites to survey their

lands and generate crop data IoT sensors may provide in-formation about crop yields rainfall pest infestation and soilnutrition which are invaluable to production and can improvefarming techniques over time Although low latency is not acritical requirement in smart farming environment manage-ment of large data sets will be a key requirement to considerMEC servers located on-site can assist high tech farmingby collecting and analyzing big data on agriculture in orderto maximize efficiency Likewise without moving everydayfarming applications to a remote cloud MEC platforms canbenefit in terms of data access synchronization storage andother overhead costs the farmer might normally incur

The use of IoT-based automated data collection and moni-toring systems in poultry houses can be used to increase workefficiency and service quality and get a deeper understandingof chicken nurturing [80] Sensing technologies can be usedin carbon dioxide and luminosity sensing these are importantparameters in large scale poultry houses Gas sensors canbe used to get all necessary information to prevent chickeninfertility due to problems such as low carbon dioxide levelsLuminosity senors can help to maintain the proper luminositylevel for optimum productivity Similar to smart farms lowlatency is not a critical requirement in smart poultry houses[80] However it is critical to manage large data sets whereon-site MEC servers can be used In addition sharing the databetween poultry houses and storing legacy data in centralizedservers are important in identifying abnormal incidents in thefarm [81] With the use of MEC poultry houses can workwith intermittent connectivity to the centralized clouds In thatcase MEC servers can temporarily hold the data until farmsare connected with the centralized clouds

H Smart Energy

The smart grid system is an Information CommunicationTechnology (ICT)-enabled energy generation transmissionand distribution network It has capabilities to continuouslysense analyze and monitor both energy flow and energytransportation infrastructure Such features are enabled byadding digital controls and enabling network monitoring andtelecommunication capabilities As a result a smart grid doesnot only provide two-way flows of electrical power but alsoenables real-time automated bidirectional flow of informa-tion Adding such smartness to the aging energy infrastructurewill foster a more efficient energy system

IoT is considered as the foundation for realizing intelli-gence capabilities in smart grid systems IoT integrates theInternet-connectivity into all kinds of grid components suchas transformers breakers switches meters relays intelligentelectronic devices capacitor banks voltage regulators cam-eras and many more These IoT devices are then used tocapture the data required to enable automations IoT-enabledsmart grids provide several benefits such as reduced capitalexpenditure optimized renewable capacity lowered mainte-nance costs and enhanced customer engagement On one handthe transformation of an electrical grid into a smart systemrequires nearly every device and piece of equipment to havebuilt-in secure interconnected intelligence On the other hand

9

an efficient system is required to manage the generated dataie transferring storing and analyzing such huge amounts ofdata which are collected from these smart devices Thereforecloud computing is a viable solution to these IoT-based smartgrids [90]

Generally smart grids are spanning over large geograph-ical areas They often confront bandwidth bottlenecks andcommunication delays due to poor network connectivity andvast number of devices generating data Thus the traditionalcentralized cloud architecture is not suitable for the domain ofthe smart grid since it relies heavily on centralized processing[91] Many delay sensitive smart grid applications such asfault detection isolation and service restoration or VoltVARoptimization cannot tolerate round trip delay to access cen-tralized cloud systems MEC is identified as the viable cloud

computing option to address these limitations MEC allowsthe computation to be performed closer to the data sourceMoreover the potential attack points for the grid is increasingwith the growth of ubiquitous sensor deployment Everysmart IoT device can be vulnerable to potential attacks MECprovides the opportunity to enforce security mechanism closerto the end devices As such even if an attacker gains accessto an endpoint device the attack gets no further informationbeyond the local network segment since MEC has capabilitiesto notice the intrusion and cease the accessibility [85]

I Industrial Internet

The Industrial Internet of Things (IIoT) also known asIndustry 40 [92] is an application of IoT in the domain ofmanufacturing IIoT incorporates numerous advanced commu-

TABLE VI The reviewed state-of-the-art MEC integration in different IoT applications

Ref Description Smar

tH

ome

Smar

tC

ity

Hea

lthca

re

IoT

Aut

omot

ive

Gam

ing

AR

VR

Ret

ail

Wea

rabl

eIo

T

Smar

tA

gric

ultu

re

Smar

tE

nerg

y

Indu

stri

alIn

tern

et

[40] Preliminary design of deploying MEC server functionalities in a smart hometo realize IoT gateway with direct M2M interaction in LTE networks

X

[41] Introduce Gateway-as-a-Service for heterogeneous IoT devices on top of thevirtualization technologies in edge computing

X X

[44] Propose an autonomic creation of MEC services to enhance QoS of videostreaming in smart cities

X

[50] Propose a semantic edge-based IoT architecture for military health services inbattlefield

X

[51] Provide a conceptual MEC based architecture for mission-critical context awarecollaboration in remote surgeries

X

[53] Describe and implement a smart e-Health gateway at the edge of the networksuitable for ubiquitous healthcare systems

X

[64] Analysis on research and engineering challenges co-existence of cloud edgecomputing and data caching strategies at the edge for vehicular networks

X

[82] Discuss the design aspects for the radio access in 5G V2X X[65] Discuss the benefits of merging MEC and mmWave technologies for 5G

applicationsX X X

[66] Propose a novel MEC-based architecture for future cellular vehicular networks X[67] Discuss the benefits of combining ICN and MEC in the context of connected

vehicle environmentsX

[52] Propose a framework for big data analytics between edge and cloud computingplatforms

X X

[74] Design and implement a fog computing based framework that support sharingand reusing contextual data across services in smart city and retail stores

X X

[83] Present a usecase of MEC for Tactile Internet based 5G gaming application X[84] A demonstration of MEC for Tactile Internet based 5G gaming application X[76] Discuss the role of MEC in 5G WIoT communication and its challenges X X[68] Propose an UAV-based IoT platform for a crowd surveillance use case X X[78] Develop and test a ubiquitous sensor network platform for crop lands automa-

tion maintenance in precision agricultureX

[70] Present a serverless edge computing architecture that enables the offloading ofmobile computation with low latency and high throughput using a mobile ARapplication

X

[85] Discuss the benefits of MEC and edge computing (EC) to enhance the securityof smart grids

X

[86] Present a method to optimize the EC based video streaming schemes forIndustrial IoT

X

[87] Present the use of edge computing to provide elastic resources and services toenable microdatabases architecture for IIoT

X

[88] Propose a fog-based communication architecture for Industry 40 applications X[73] Describe research directions and enablers of wireless interconnected VR

systemsX

[89] Design an optimization framework for VRAR communication via small-cellcooperation

X

10

nication and automation technologies such as M2M commu-nication machine learning and big data analytics to improveintelligence and the connectivity [93] For instance IIoTnetworks can connect all of the employees data and processesfrom the factory floor and forward them to the executiveoffices Thus decision makers or employees can create a fulland accurate view of their manufacturing process by usingIIoT network hence improving their ability to make moreinformed decisions IIoT also helps the exploitation as well asimplementation of new intelligent technologies to acceleratethe innovation and transformation of the factory workforce[92]

Primarily IIoT is seen as a way to improve operationalefficiency However IIoT provides a wide range of otherbenefits such as improving connectivity efficiency scalabilitytime savings as well as cost savings for manufacturing pro-cesses with the maximum use of smart machines [92] [94] Ingeneral these smart machines operate with higher accuracygreater efficiency and constant working capabilities than hu-mans [95] Thus IIoT has great potential for improving qualitycontrol sustainability and overall supply chain efficiency

MEC will play a vital role in enabling future IIoT applica-tions [96] by addressing the shortcomings of M2M communi-cation (eg latency resilience cost peer-to-peer connectivitysecurity) in IIoT domain [97] [98] Current market trendsalready show that edge computing will represent many im-plementation scenarios for IIoT For instance real-time edgeanalytics and enhanced edge security are two key drivers inthe creation of new IIoT deployments Thus the addition ofMEC in IIoT networks will fuel the evolution of IIoT as wellas create new business applications [99]

One way to optimize the use of conventional edge com-puting in video streaming schemes for IIoT is presented in[86] By using machine learning algorithms edge computingcan process the sensor data before transmitting to the cloudThis mitigates against the degradation of service quality of thevideo streaming Aggregation of all the sensor data to a singledata center increases latency and raises performance concernsin IIoT domain In order to solve this issue a microdatabasearchitecture is proposed for the Industrial Internet [87] It holdsthe data close to the industrial processes but also makes itavailable near the applications that can benefit from the dataEdge computing also provides elastic resources and services toenable micro-database architecture [87] A fog-based commu-nication architecture for Industry 40 applications is proposedin [88] This approach will substantially minimize the energyconsumption of the IoT nodes Edge computational capabilitiesare further used to predict future data measurements andreduce the throughput from IoT devices to the control unit

III TECHNICAL ASPECTS OF MEC ENABLED IOT

To realize the MEC exploitation for IoT applications thekey value propositions are mostly seen from the technicalparameters such as scalability communication computationoffloading and resource allocation mobility management se-curity privacy and trust management This section describesthe state-of-the-art of each of these technical parameters hence

giving a clear background against which the benefits of MECcan be envisioned

A Scalability

1) Requirements When it comes to actual deployment ofMEC platform for IoT systems scalability is a key factorto consider The compatibility of MEC servers to multiplenetwork environments is one of the factors that will driveits large scale adoption in future networks [100] The IoTenvironment will consist of hundreds of billions of sensors ac-tuators Radio-Frequency Identification (RFID)-tagged objectssoftware vehicles and embedded systems all interconnectedin a huge network of cyber-physical systems At a utility scaleconsideration these devices will be working in close collab-oration to deliver the expected services in technologies likethe smart grids virtual power plants smart homes intelligenttransportation and smart cities That being said the role ofscalability to the realization of such a hyper-connected IoTenvironment becomes more obvious The IoT environment willrequire a dynamic range of capabilities in the network space ifsuch large numbers of devices are to be supported effectively

2) Related work Currently MEC servers have been con-firmed to be compatible with LTE macro base station (eN-odeB) sites 3G Radio Network Controller (RNC) site multi-Radio Access Technology (RAT) cell aggregation site andat the edge of the core network [2] Such multi-RAT cellaggregation schemes can be implemented indoor or outdoorsettings depending on the requirements This invariably en-ables MEC to be applied to many different possible scenariosThe larger the deployment scenarios for MEC the more therange of capabilities it can handle this also translates to higherscalability for MEC-enable technologies like IoT

Designing an edge cloud network implies that an optimallocation for citing the cloud facility is first determined In[105] authors present a design optimization scheme for theMEC architecture based on link-path formulation supportedby heuristics in order to optimize the computation time forthe scheme In this approach consideration is given to bothusers and VMs mobility Hence an optimal point to installthe MEC server is determined through a tread-off betweeninstallation cost and the quality of service to be deliveredTable VII compares the reviewed state-of-the-art scalabilityfeature in MEC enabled IoT

B Communication

1) Requirements There are three main categories for thecommunication concerns about MEC [106] Wireless accesswhile offloading to the mobile edge host Backhaul accesswhile offloading to a remote cloud server Communicationamong IoT devices mobile edge host and remote cloudservers when they collaboratively execute multiple jobs Thefirst and the second categories are the most renowned on behalfof the MEC servers which are the small scale data centersdeployed by the network operators and can be co-located withthe Wireless Access Points (WAPs) In the IoT supportiveMEC systems the consumer devices may communicate with

11

TABLE VII Comparison of the reviewed state-of-the-art scalability feature in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Addressing Search

[101] Discusses the challenges in searching imposed by the burgeoning field of IoT General IoT X[94] Examines a variety of popular and innovative IoT solutions in terms of context-aware

technology perspectives to serve as a conceptual framework for context-aware productdevelopment and research in the IoT paradigm

General IoT solu-tions

X

[102] Proposes an innovative distributed architecture combining machine-to-machine industry-mature protocols (ie MQTT and CoAP) in an original way to enhance the scalabilityof gateways for the efficient IoT-cloud integration

IoT cloud inte-gration

X X

[103] Studies an implementation of edge computing which exploits transparent computing tobuild scalable IoT platforms using transparent computing

Wearable IoT X

[104] Introduces a lightweight edge gateway for the IoT architecture using container-basedvirtualization techniques

General IoT X

the MEC servers either directly or with the support of neigh-boring devices using Device-to-Device (D2D) communicationFor the third category WAPs enable access to the remote datacenters in the central cloud through backhaul links

In order to reap the maximum advantage of computationoffloading leveraged at the edge servers MEC systems needefficient communication channels Unlike the wired connec-tions in the conventional grid computing and cloud computingthe wireless access links between the mobile devices andcloud computing resources in the edge computing paradigmcan be unstable Sudden service outages may occur with theinterruption of access links The inherent challenges withwireless communication channels like multi-path fading in-terference and spectrum shortage should always be taken intoaccount for the design of MEC systems to seamlessly integratecomputation offloading and radio resource management [32]Moreover both wireless and backhaul access links have lim-ited capacities which should be properly shared among mobiledevices in a similar way as sharing the computing resources ofthe MEC server Hence having a cooperative scheme for thejoint allocation of communication and computation resourcesis important for the successful deployment of MEC [106]Redesigning both communication and networking protocolsto integrate communication infrastructures in MEC and IoTsystems is a challenging task The key focus should be onimproving the computation efficiency with respect to datatransmission

Another major requirement is to maintain interoperabilitywhile addressing heterogeneous communication technologiesthat have to be utilized in IoT and MEC paradigms in 5GThere are plenty of radio technologies that facilitate IoT Low-Power Wide Area Networks (LPWANs) (eg WCDMA LTEnarrowband IoT (NB-IoT) Wi-Fi Bluetooth Zigbee SIGFOXand LoRA) The choice of these LPWAN technologies maycreate trade-offs among signal strength operational rangethroughput and power consumption With the arrival of 5Gthe convergence of these communication technologies needsto be achieved since one network will not be fitting based onthose trade-offs

2) Related work Recently Fog-Radio Access Network (F-RAN) was introduced by Peng et al to consolidate the hetero-geneous networks into a single network architecture with 5Geven though they do not operate in the same bands to gain highspectral and operating and energy efficiency [107] Well known

Cloud Radio Access Network (C-RAN) architecture can per-form cooperative transmission across multiple edge nodes withcentralized cloud computing servers via fronthaul links [108]Although C-RAN provides high spectral efficiencies due tothe enhanced interference management capabilities with thecentralized baseband processing at the cloud it has potentiallylarge latencies F-RAN is proposed for 5G MEC deploymentsas an advanced socially aware mobile networking architectureto provide high spectral efficiency while maintaining highenergy efficiency and low latency [107] [108] Precodingdesign resource block allocation user scheduling and cellassociation are jointly designed for radio resource allocation inF-RANs in order to optimize spectral and energy efficienciesand latency performances [109] In [110] Rimal et al pro-pose a unified Time-Division Multiple Access (TDMA) basedresource management scheme for offloading traffic over Fiber-enabled Wireless (FiWi) access networks

In the envisioned 5G systems and MEC architecture bothbackhaul and wireless access links can be facilitated bymillimeter-Wave (mmW) spectrum [111] The use of mmWspectrum will enable high data rate access to MEC function-alities with low latency On the other hand MEC provideslocal computation power usefully for optimizing the perfor-mance of mmW communications In [112] [113] the authorsaddress the joint optimization of communicationcomputationresources with mmW communication They have taken the ad-vantage of blocking probabilities by considering intermittencyof mmW multi-link communications

An open source LPWAN infrastructure called OpenChirpis discussed in [114] OpenChirp which is developed usingLoRWAN allows multiple users to provision and to managebattery-powered transducers across large areas like campusesindustrial zones or cities As pointed out in [30] [115]SDN plays a vital role in improving MEC type technologiesby removing the technical shortcomings in edge computingimplementations The authors summarize the work performedfor implementing MEC based on NFV and SDN where theSDN controller manages the communication between MECservers which form a data center at the edge Table VIIIsummarizes the reviewed state-of-the-art communication is-sues and solutions in MEC enabled IoT

12

TABLE VIII Comparison of the reviewed state-of-the-art communication issues and solutions in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Commnetworkarchitecture

Commresourceallocation

[109] Performance analysis of radio resource allocation in F-RANs for edge cache andadaptive model selection to improve spectral efficiency and energy efficiency

Low latency andhigh reliability

X

[112][113]

Use of mmWave spectrum for high data rate access to MEC servers and backhaul links Low latency andhigh reliability

X

[114] An open source LPWAN infrastructure which allows multiple users to provision andmanage battery-powered transducers across large areas

LPWANinfrastructure

X

[42] A virtualized edge computing architecture with a proxy VM migration scheme tominimize traffic in the core network

IoT big datastreams

X

[115] Proposed network architecture includes multi-interface wireless access network(eg FiWi) heterogeneous backhauling distributed cloudlets hierarchical structure ofa cloudlet and the SDN based mobile core network

IoT big datastreams

X

[110] A novel unified resource management scheme for Ethernet-based FiWi networks thatjointly allocates bandwidth for transmissions of both conventional broadband traffic andMEC data in a TDMA fashion

Mission-criticalIoT

X

[116] Introduce Mobile-IoT-Federation-as-a-Service (MIFaaS) to enable dynamic cooperationamong privatepublic local clouds of IoT devices at the edge of the cellular infrastructureThe selection of the best configuration of federated IoT cloud platforms are modeledas a coalition formation problem

Cellular IoT X

[117] Allocation of radio resources in a joint LTE and NB-IoT system based of MIFaasparadigm [116] Discovered that in handling high-end IoT data traffic a combinationbetween NB-IoT and LTE is essential in providing the needed high data rate and lowlatency

Mission-criticalIoT

X

[118] Integration of D2D communications into edge computing environment reduce transmis-sion delay and traffic load across the network

Mission-criticalIoT

X

[119] Use the theories of stochastic geometry queueing and parallel computing for provi-sioning and planning MEC networks

Communicationlatency

X

C Computation Offloading and Resource Allocation

1) Requirements Computation offloading is the mostprominent and widely discussed feature of MEC that empow-ers resource-constrained IoT devices with augmented com-putational capabilities [29] [33] This will not only prolongthe battery life of the IoT sensor nodes but also reduceE2E latency needed to run sophisticated applications In thefirst place UE has to decide whether to execute the rela-tively simple tasks locally or offload to the MEC servers(ie task model for binary offloading) [32] Secondly thedecision of computation offloading to the MEC servers can beperformed fully or partially In the partial offloading a subsetof computations is executed locally while the rest is offloadedto the MEC server by considering several factors such asusers or application preferences (eg application buffer state)radio and backhaul connections quality (ie between UEand MEC servers) UE capabilities or cloud capabilities andavailability [29]

The sole objective of the offloading policies need to be theminimization of execution delay Other critical concerns areto define the dependency of offloadable components of theapplications based on their ability to partition data (eg real-time user input has to be processed at UE without offloading)and to predict the execution time of multiple tasks Theexecution order or routines have to be carefully formulatedsince certain outcomes can be the inputs of other tasks Aspointed out in [32] the task models for partial offloading canbe represented by task-call graphs with sequential paralleland general dependencies

Although in MEC computation offloading enables power-ful cloud services at the edge level the insufficient batteryenergy at the tiny IoT devices may incur new challenges In

applications like IoT surveillance or remote asset managementthe nodes are typically hard to reach Those applications mayalso require to offload data more frequently in small chunksby consuming more energy Therefore it is necessary to con-sider not only the trade-off between energy consumption andexecution delay in both full and partial offloading scenarios inMEC but also the trade-off between computation energy andtransmission energy consumption in order to extend batterylife

The joint computation and communication resource alloca-tion should be properly addressed in order to get the maximumutilization of available resources Single MEC server will beallocated for the applications which cannot be partitionedThe resources in multiple MEC servers are allocated for theoffloaded applications that can be split into several parts Whena job arrives at the MEC server if there are enough resourcesthe scheduler has to allocate the VM for further processingIf there are no sufficient computation resources it delegatesthe task to the centralized cloud MEC servers also have toallocate computation and communication resources for userapplication jobs and MEC service jobs User mobility networktopology network scalability and load balancing are someother factors to be considered in order to define fare resourceutilization policies on MEC servers Specifically when IoTgateways share limited bandwidth among multiple IoT deviceswhich can handle video audio or bio-medical signals theallocation of bandwidth will become challenging [120] Thelow power wireless technologies (eg BLE ZigBee lowpower Wi-Fi and LPWAN standards like LoRA or SigFox)used in IoT networks have limited bandwidth When the IoTdevices access the MEC server which is acting as the IoTgateway they have to utilize either of those low-power wireless

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

6

TABLE V MEC and IoT benefits for each application

Required characteristicsof MEC and IoT

Description Smar

tho

me

Smar

tci

ty

Rem

ote

surg

ery

Rem

ote

heal

thco

nsul

tanc

y

Aut

onom

ous

vehi

cles

Aug

men

ted

Rea

lity

(AR

)

Vir

tual

Rea

lity

(VR

)

Gam

ing

Ret

ail

Wea

rabl

eIo

T

Farm

ing

Smar

ten

ergy

Indu

stri

alIn

tern

et

Low Latency Optimize to process a very high volume ofdata messages with minimal delay

X X X X X X X X X X

Increased Bandwidth Ability move a large set amount of datarapidly

X X X X X X X X X X X X

Content Awareness Adaptation of network characteristics ac-cording the local services requirements

X X X X X X X X X X

Low power devices Support for low power devices which haslimited transmission powers

X X X X X X X

Fixed wireless support Operation of wireless systems used to con-nect two fixed locations with a wireless link

X X X X X X X X X X

Fast inter-RAT handoff Speed up the handover takes place betweendifferent RATs

X X X X X X X X

Caching Keeping frequently accessed information ina location close to the requester

X X X X X X X

Edge Analytics An automated analytical computation is per-formed on data at a sensor network switchor other device instead of waiting for thedata to be sent back to a centralized datastore

X X X X X X X X X

Application virtualizationbetween edge and cloud

On demand application and service migra-tion from centralized cloud to the edgecloud

X X X X X X X X X X X

Private or local network Limit the communication and data ex-changes to a certain network segment

X X X X X X X X X X X X

Security Provide localized security X X X X X X XPrivacy Provide localized Privacy X X X X X X XFast Mobility Enable the ability to move or be moved fast

within the network or network coverablearea

X X X X X X X X

B Healthcare

Mobile health and telemedicine are identified as importantuse cases of 5G Wearable low power IoT medical sensorsfor monitoring health related data and tracking records arenow popular in public healthcare facilities [46] Although IoTtechnologies are widely adopted in the health sector [47]their performance goals will not be achievable without edgecomputing solutions like MEC [37] [48] [49] For instancehumanoid robots sitting next to an elderly person may needtactile feedback in 1ms latency for his or her care taking ser-vices Mission critical use cases like remote surgeries requireultra-low latency uninterrupted communication links andcollaborations among surgeons present in different locationsRemote patient monitoring is another use case which enablesconsultants in major cities to interact with patients residingfar away from the medical facility The frequent updates ofhealth records for an elderly person or someone with a chronicdisease needs to proceed ubiquitously and securely With suchpotential use cases and scenarios the role of MEC in healthand social assistance industries becomes more evident [37]

Some research works have already been published aboutthe cooperation between edge computing and IoT in thehealthcare sector In [50] authors describe a military health-

care service platform based on hierarchical IoT architectureand a semantic edge network model The hierarchical IoTarchitecture can collect the vital health parameters of thesoldiers their weapon status as well as their geographicallocations The control center of the battlefield performs therole of edge component which can process and store largeamount of health data sent over an SDN-based network Thepreliminary network architecture proposed in [51] providesreal-time context-aware collaboration for remote robotic tele-surgeries Big data analytics performed by edge computingare also important in e-Healthcare applications [52] In [53]Rahmani et al introduces the smart gateway concept foran IoT-based remote health monitoring system Here they exploit edge computing nodes to update the centralized cloudbased on the medical data generated by the IoT sensors Theirgeo-distributed network of smart e-Health gateways provideslocal data processing for real-time notification for medicalpractitioners secure and privacy preserved data gathering pa-tientsrsquo mobility network interoperability and energy efficientcommunication

C Autonomous VehiclesIoT Automotive5G is a key enabler of V2X (Vehicle to Everything) concept

which covers Vehicle to Vehicle (V2V) vehicle to infras-

7

tructure vehicle to device vehicle to pedestrian vehicle tohome and vehicle to grid [54] In the context of IoT Au-tomotive V2X requires critical communication infrastructurewhere reliability and ultra low latency are crucial factors [55]Use cases in these categories include autonomous and semi-autonomous driving vehicle maintenance and in vehicle info-tainment In order to operate an efficient and reliable vehicularnetwork several features have to be improved these includereal-time traffic monitoring [56] [57] continuous sensing invehicles [58] [59] support for Infotainment applications [60]and improved security [61] However these features cannot beserved by current mobile networks [62] In this vein upcoming5G mobile systems are expected to offer a higher level of flexi-bility leveraging the emerging technologies related to networksoftwarization [63] In this context V2X combined with MECprovides a viable and cost-effective solution that can acceleratedevelopment of V2X and IoT automotive systems [64]

It is important to improve the performance of RAN tech-nologies to enable IoT automatization MEC will play a vitalrole here also For instance MEC technologies may fulfillthe latency reliability and throughput requirements in V2Xchannel modeling of mmWave communication [65] Moreoverthe placement of the MEC server within the RAN providesflexible network services for the vehicle and to efficientlycontrol the radio network resources [66] It is also possibleto design a time-predicted handover mechanism for vehiclesby leveraging road side information at MEC server in order tomeet the demand for high mobility and reliability in vehicularnetworks [66]

In addition ICN-MEC integration can also tackle existingtechnical challenges such as massive mobility of vehicles scal-ability deployment strategies service orchestration massivedata handling fast big data processing as well as ensuringsecurity and privacy [67]

Unmanned aerial vehicles (UAVs) or drones are anothertype of autonomous vehicles which are capable of sensingits environment and navigating without human inputs UAVuse cases include but not limited to public safety smart agri-culture surveillance and environmental monitoring [68] Inorder to maximize the flight time the UAV battery life shouldbe essentially conserved by minimizing the overhead onboardWhen the required processing power exceeds the availableresources on UAV the application data can be offloaded toMEC Accompanying the advanced RATs MEC will facilitatethe offloading process from UAV due to its expected widedeployment in the network [68]

D Gaming AR and VR

Mixed reality (MR) combines virtual reality (VR) andaugmented reality (AR) technologies thereby enabling humansto interact more naturally with the virtual worlds based ondata aggregated by IoT devices [69] With IoT AR tech-nologies are able to benefit directly from the high end inter-connection of objects that characterizes the IoT environmentthrough which users can extend their interactions from thereal world to the virtual world [2] [70] Convergence ofVR and IoT can occur in many ways such as telepresence

tourism industry smart transportation networks and roboticassisted surgeries Exclusive AR and VR experiences withthe delivery of 360 navigable videos will be offered byenhanced mobile broadband connections with low latencyand high reliability for mission-critical services With present-day network standards this might be impossible to achievehowever with the predicted characteristics of 5G such as20 Gbps peak data rate and 1 ms round-trip over-the-airlatency this becomes more easily achievable As identified byETSI MEC will be an ideal solution for low-latency offloadservices in AR and VR applications that combine computergenerated data with physical reality [71] While operatingVR devices over wireless links and deploying the VR controlcenter at MEC server the tracking accuracy can be increasedwith round trip latency of 1 ms and high reliability [72]Migrating computationally intensive tasks to edge servers willincrease the computational capacity of VR devices and savetheir battery-life Furthermore MEC will allow VR devices toaccess cloud resources in an on-demand fashion [73]

MEC platforms provide high capacity and low latencywireless coverage for large venues like stadiums or smartcities with a massive density of users to enjoy the AR andVR experience For instance inside a smart building with anetwork of cameras obtaining raw video frames and preparingthe processed frames for display can be performed locallywith the help of edge computing Furthermore tracking thelocal position of the user or object building a model of theenvironment and identifying known objects in the environ-ment can be offloaded to the edge cloud Similarly in orderto get absolute experience of VR glasses the response timeshould be extremely low When the user moves his head hemay experience delay if the glasses need to access remotedata centers Therefore the expected interaction time betweenmachines and humans needs to be less than 1ms When thelatency of a VR application is more than 1ms the user willexperience cyber sickness which will be interrupting the realVR experience MEC servers in the nearest proximity will beable to serve such applications with ultra low latency Futuregames will be played beyond the entertainment purposes ontop of VR and AR applications which would require theminimum possible latency Pokmon Go and Ingress are twoexamples of successful games that combine AR and sensorinformation such as user location

E Retail

The second largest MEC use case is expected to be in the re-tail businesses [37] Currently IoT has dominated retail marketapplications in many ways including digital signage supplychain management intelligent payment solutions smart vend-ing machines shelves doors resource management stream-ing and safety The high class retail stores which use facialrecognition systems need high definition cameras that generatehuge volumes of data requiring powerful servers within thepremises Therefore the on-site MEC servers will assist toprocess these kind of large data sets produced by IoT devicesin a retail market Big data analytics in shopping centerscan further exploit the collaborative processing between edge

8

and cloud computing [52] Installation of MEC in a retailmarket also provides high speed mobile coverage throughoutthe store WiFi access points that are maintained per store canbe connected to the MEC server to provide WiFi connectivityfor store customers as needed The enabling of MEC willalso omit load balancing Wi-Fi controllers or policy enginesrequired in the wide area networks in the store Althoughnot many academic published research works are explicitlyfocusing on MEC and IoT [74] they have become enormouslyreputed and commercialized technologies in the industry andthe business sectors

F Wearable IoT (WIoT)

During the previous years wearable technology has evolvedtremendously from walkman to step trackers smart watchesto smart glasses The development of low power wirelesstechnologies such as BLE (Bluetooth Low Energy) fuels thedevelopment of wearable devices Present-day wearables spanfrom low-end devices such as health and fitness trackers tohigh-end devices such as VRAR helmets and smart watchesIt is expected that wearables will become the worlds best-selling consumer electronics product after smartphones witha global availability of more than 929 million devices by2021 [75] With the new application domains and enablingservices wearable devices will demand more sophisticatedcommunication infrastructures For instance VRAR wear-ables are demanding gigabits throughput network connectivityto run their applications On the other hand dense deploymentof wearable devices in smart cities will increase the networktraffic on communication networks Thus the next generationcommunication networks should be able to provide the gigabitexperience for the anticipated ultra dense wearable devices[76]

Although cloud computing has enabled wide range of newnetworking services it cannot alone fulfill the upcoming re-quirements for the future wearable ecosystem Mainly the cen-tralized cloud data centers fails due to long End-to-End (E2E)latency Delay-sensitive wearable applications such as VRperceptual stability requires ultra low delay In this contextMEC has the potential to solve the limitations in current cloudbased systems by combining cloud and MEC infrastructuresThis will enable providers deploy storage computing andcaching capabilities in close proximity with such wearabledevices [76]

G IoT in Mechanized Agriculture

In order to meet the demands for future food productionthe agricultural sector will require some major evolution whereIoT will be integrated in various production management andanalytical processes [77] [78] The present-day agriculturalsector has been slow to adopting the emerging Machine-to-Machine (M2M) and IoT technologies when compared withother sectors like smart cities and the medical fields [79]

Precision farming and smart agriculture can be achievedusing autonomous vehicles (tractors) remote monitoring andreal-time analytics It is reported that farmers are increasinglyturning to agricultural drones and satellites to survey their

lands and generate crop data IoT sensors may provide in-formation about crop yields rainfall pest infestation and soilnutrition which are invaluable to production and can improvefarming techniques over time Although low latency is not acritical requirement in smart farming environment manage-ment of large data sets will be a key requirement to considerMEC servers located on-site can assist high tech farmingby collecting and analyzing big data on agriculture in orderto maximize efficiency Likewise without moving everydayfarming applications to a remote cloud MEC platforms canbenefit in terms of data access synchronization storage andother overhead costs the farmer might normally incur

The use of IoT-based automated data collection and moni-toring systems in poultry houses can be used to increase workefficiency and service quality and get a deeper understandingof chicken nurturing [80] Sensing technologies can be usedin carbon dioxide and luminosity sensing these are importantparameters in large scale poultry houses Gas sensors canbe used to get all necessary information to prevent chickeninfertility due to problems such as low carbon dioxide levelsLuminosity senors can help to maintain the proper luminositylevel for optimum productivity Similar to smart farms lowlatency is not a critical requirement in smart poultry houses[80] However it is critical to manage large data sets whereon-site MEC servers can be used In addition sharing the databetween poultry houses and storing legacy data in centralizedservers are important in identifying abnormal incidents in thefarm [81] With the use of MEC poultry houses can workwith intermittent connectivity to the centralized clouds In thatcase MEC servers can temporarily hold the data until farmsare connected with the centralized clouds

H Smart Energy

The smart grid system is an Information CommunicationTechnology (ICT)-enabled energy generation transmissionand distribution network It has capabilities to continuouslysense analyze and monitor both energy flow and energytransportation infrastructure Such features are enabled byadding digital controls and enabling network monitoring andtelecommunication capabilities As a result a smart grid doesnot only provide two-way flows of electrical power but alsoenables real-time automated bidirectional flow of informa-tion Adding such smartness to the aging energy infrastructurewill foster a more efficient energy system

IoT is considered as the foundation for realizing intelli-gence capabilities in smart grid systems IoT integrates theInternet-connectivity into all kinds of grid components suchas transformers breakers switches meters relays intelligentelectronic devices capacitor banks voltage regulators cam-eras and many more These IoT devices are then used tocapture the data required to enable automations IoT-enabledsmart grids provide several benefits such as reduced capitalexpenditure optimized renewable capacity lowered mainte-nance costs and enhanced customer engagement On one handthe transformation of an electrical grid into a smart systemrequires nearly every device and piece of equipment to havebuilt-in secure interconnected intelligence On the other hand

9

an efficient system is required to manage the generated dataie transferring storing and analyzing such huge amounts ofdata which are collected from these smart devices Thereforecloud computing is a viable solution to these IoT-based smartgrids [90]

Generally smart grids are spanning over large geograph-ical areas They often confront bandwidth bottlenecks andcommunication delays due to poor network connectivity andvast number of devices generating data Thus the traditionalcentralized cloud architecture is not suitable for the domain ofthe smart grid since it relies heavily on centralized processing[91] Many delay sensitive smart grid applications such asfault detection isolation and service restoration or VoltVARoptimization cannot tolerate round trip delay to access cen-tralized cloud systems MEC is identified as the viable cloud

computing option to address these limitations MEC allowsthe computation to be performed closer to the data sourceMoreover the potential attack points for the grid is increasingwith the growth of ubiquitous sensor deployment Everysmart IoT device can be vulnerable to potential attacks MECprovides the opportunity to enforce security mechanism closerto the end devices As such even if an attacker gains accessto an endpoint device the attack gets no further informationbeyond the local network segment since MEC has capabilitiesto notice the intrusion and cease the accessibility [85]

I Industrial Internet

The Industrial Internet of Things (IIoT) also known asIndustry 40 [92] is an application of IoT in the domain ofmanufacturing IIoT incorporates numerous advanced commu-

TABLE VI The reviewed state-of-the-art MEC integration in different IoT applications

Ref Description Smar

tH

ome

Smar

tC

ity

Hea

lthca

re

IoT

Aut

omot

ive

Gam

ing

AR

VR

Ret

ail

Wea

rabl

eIo

T

Smar

tA

gric

ultu

re

Smar

tE

nerg

y

Indu

stri

alIn

tern

et

[40] Preliminary design of deploying MEC server functionalities in a smart hometo realize IoT gateway with direct M2M interaction in LTE networks

X

[41] Introduce Gateway-as-a-Service for heterogeneous IoT devices on top of thevirtualization technologies in edge computing

X X

[44] Propose an autonomic creation of MEC services to enhance QoS of videostreaming in smart cities

X

[50] Propose a semantic edge-based IoT architecture for military health services inbattlefield

X

[51] Provide a conceptual MEC based architecture for mission-critical context awarecollaboration in remote surgeries

X

[53] Describe and implement a smart e-Health gateway at the edge of the networksuitable for ubiquitous healthcare systems

X

[64] Analysis on research and engineering challenges co-existence of cloud edgecomputing and data caching strategies at the edge for vehicular networks

X

[82] Discuss the design aspects for the radio access in 5G V2X X[65] Discuss the benefits of merging MEC and mmWave technologies for 5G

applicationsX X X

[66] Propose a novel MEC-based architecture for future cellular vehicular networks X[67] Discuss the benefits of combining ICN and MEC in the context of connected

vehicle environmentsX

[52] Propose a framework for big data analytics between edge and cloud computingplatforms

X X

[74] Design and implement a fog computing based framework that support sharingand reusing contextual data across services in smart city and retail stores

X X

[83] Present a usecase of MEC for Tactile Internet based 5G gaming application X[84] A demonstration of MEC for Tactile Internet based 5G gaming application X[76] Discuss the role of MEC in 5G WIoT communication and its challenges X X[68] Propose an UAV-based IoT platform for a crowd surveillance use case X X[78] Develop and test a ubiquitous sensor network platform for crop lands automa-

tion maintenance in precision agricultureX

[70] Present a serverless edge computing architecture that enables the offloading ofmobile computation with low latency and high throughput using a mobile ARapplication

X

[85] Discuss the benefits of MEC and edge computing (EC) to enhance the securityof smart grids

X

[86] Present a method to optimize the EC based video streaming schemes forIndustrial IoT

X

[87] Present the use of edge computing to provide elastic resources and services toenable microdatabases architecture for IIoT

X

[88] Propose a fog-based communication architecture for Industry 40 applications X[73] Describe research directions and enablers of wireless interconnected VR

systemsX

[89] Design an optimization framework for VRAR communication via small-cellcooperation

X

10

nication and automation technologies such as M2M commu-nication machine learning and big data analytics to improveintelligence and the connectivity [93] For instance IIoTnetworks can connect all of the employees data and processesfrom the factory floor and forward them to the executiveoffices Thus decision makers or employees can create a fulland accurate view of their manufacturing process by usingIIoT network hence improving their ability to make moreinformed decisions IIoT also helps the exploitation as well asimplementation of new intelligent technologies to acceleratethe innovation and transformation of the factory workforce[92]

Primarily IIoT is seen as a way to improve operationalefficiency However IIoT provides a wide range of otherbenefits such as improving connectivity efficiency scalabilitytime savings as well as cost savings for manufacturing pro-cesses with the maximum use of smart machines [92] [94] Ingeneral these smart machines operate with higher accuracygreater efficiency and constant working capabilities than hu-mans [95] Thus IIoT has great potential for improving qualitycontrol sustainability and overall supply chain efficiency

MEC will play a vital role in enabling future IIoT applica-tions [96] by addressing the shortcomings of M2M communi-cation (eg latency resilience cost peer-to-peer connectivitysecurity) in IIoT domain [97] [98] Current market trendsalready show that edge computing will represent many im-plementation scenarios for IIoT For instance real-time edgeanalytics and enhanced edge security are two key drivers inthe creation of new IIoT deployments Thus the addition ofMEC in IIoT networks will fuel the evolution of IIoT as wellas create new business applications [99]

One way to optimize the use of conventional edge com-puting in video streaming schemes for IIoT is presented in[86] By using machine learning algorithms edge computingcan process the sensor data before transmitting to the cloudThis mitigates against the degradation of service quality of thevideo streaming Aggregation of all the sensor data to a singledata center increases latency and raises performance concernsin IIoT domain In order to solve this issue a microdatabasearchitecture is proposed for the Industrial Internet [87] It holdsthe data close to the industrial processes but also makes itavailable near the applications that can benefit from the dataEdge computing also provides elastic resources and services toenable micro-database architecture [87] A fog-based commu-nication architecture for Industry 40 applications is proposedin [88] This approach will substantially minimize the energyconsumption of the IoT nodes Edge computational capabilitiesare further used to predict future data measurements andreduce the throughput from IoT devices to the control unit

III TECHNICAL ASPECTS OF MEC ENABLED IOT

To realize the MEC exploitation for IoT applications thekey value propositions are mostly seen from the technicalparameters such as scalability communication computationoffloading and resource allocation mobility management se-curity privacy and trust management This section describesthe state-of-the-art of each of these technical parameters hence

giving a clear background against which the benefits of MECcan be envisioned

A Scalability

1) Requirements When it comes to actual deployment ofMEC platform for IoT systems scalability is a key factorto consider The compatibility of MEC servers to multiplenetwork environments is one of the factors that will driveits large scale adoption in future networks [100] The IoTenvironment will consist of hundreds of billions of sensors ac-tuators Radio-Frequency Identification (RFID)-tagged objectssoftware vehicles and embedded systems all interconnectedin a huge network of cyber-physical systems At a utility scaleconsideration these devices will be working in close collab-oration to deliver the expected services in technologies likethe smart grids virtual power plants smart homes intelligenttransportation and smart cities That being said the role ofscalability to the realization of such a hyper-connected IoTenvironment becomes more obvious The IoT environment willrequire a dynamic range of capabilities in the network space ifsuch large numbers of devices are to be supported effectively

2) Related work Currently MEC servers have been con-firmed to be compatible with LTE macro base station (eN-odeB) sites 3G Radio Network Controller (RNC) site multi-Radio Access Technology (RAT) cell aggregation site andat the edge of the core network [2] Such multi-RAT cellaggregation schemes can be implemented indoor or outdoorsettings depending on the requirements This invariably en-ables MEC to be applied to many different possible scenariosThe larger the deployment scenarios for MEC the more therange of capabilities it can handle this also translates to higherscalability for MEC-enable technologies like IoT

Designing an edge cloud network implies that an optimallocation for citing the cloud facility is first determined In[105] authors present a design optimization scheme for theMEC architecture based on link-path formulation supportedby heuristics in order to optimize the computation time forthe scheme In this approach consideration is given to bothusers and VMs mobility Hence an optimal point to installthe MEC server is determined through a tread-off betweeninstallation cost and the quality of service to be deliveredTable VII compares the reviewed state-of-the-art scalabilityfeature in MEC enabled IoT

B Communication

1) Requirements There are three main categories for thecommunication concerns about MEC [106] Wireless accesswhile offloading to the mobile edge host Backhaul accesswhile offloading to a remote cloud server Communicationamong IoT devices mobile edge host and remote cloudservers when they collaboratively execute multiple jobs Thefirst and the second categories are the most renowned on behalfof the MEC servers which are the small scale data centersdeployed by the network operators and can be co-located withthe Wireless Access Points (WAPs) In the IoT supportiveMEC systems the consumer devices may communicate with

11

TABLE VII Comparison of the reviewed state-of-the-art scalability feature in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Addressing Search

[101] Discusses the challenges in searching imposed by the burgeoning field of IoT General IoT X[94] Examines a variety of popular and innovative IoT solutions in terms of context-aware

technology perspectives to serve as a conceptual framework for context-aware productdevelopment and research in the IoT paradigm

General IoT solu-tions

X

[102] Proposes an innovative distributed architecture combining machine-to-machine industry-mature protocols (ie MQTT and CoAP) in an original way to enhance the scalabilityof gateways for the efficient IoT-cloud integration

IoT cloud inte-gration

X X

[103] Studies an implementation of edge computing which exploits transparent computing tobuild scalable IoT platforms using transparent computing

Wearable IoT X

[104] Introduces a lightweight edge gateway for the IoT architecture using container-basedvirtualization techniques

General IoT X

the MEC servers either directly or with the support of neigh-boring devices using Device-to-Device (D2D) communicationFor the third category WAPs enable access to the remote datacenters in the central cloud through backhaul links

In order to reap the maximum advantage of computationoffloading leveraged at the edge servers MEC systems needefficient communication channels Unlike the wired connec-tions in the conventional grid computing and cloud computingthe wireless access links between the mobile devices andcloud computing resources in the edge computing paradigmcan be unstable Sudden service outages may occur with theinterruption of access links The inherent challenges withwireless communication channels like multi-path fading in-terference and spectrum shortage should always be taken intoaccount for the design of MEC systems to seamlessly integratecomputation offloading and radio resource management [32]Moreover both wireless and backhaul access links have lim-ited capacities which should be properly shared among mobiledevices in a similar way as sharing the computing resources ofthe MEC server Hence having a cooperative scheme for thejoint allocation of communication and computation resourcesis important for the successful deployment of MEC [106]Redesigning both communication and networking protocolsto integrate communication infrastructures in MEC and IoTsystems is a challenging task The key focus should be onimproving the computation efficiency with respect to datatransmission

Another major requirement is to maintain interoperabilitywhile addressing heterogeneous communication technologiesthat have to be utilized in IoT and MEC paradigms in 5GThere are plenty of radio technologies that facilitate IoT Low-Power Wide Area Networks (LPWANs) (eg WCDMA LTEnarrowband IoT (NB-IoT) Wi-Fi Bluetooth Zigbee SIGFOXand LoRA) The choice of these LPWAN technologies maycreate trade-offs among signal strength operational rangethroughput and power consumption With the arrival of 5Gthe convergence of these communication technologies needsto be achieved since one network will not be fitting based onthose trade-offs

2) Related work Recently Fog-Radio Access Network (F-RAN) was introduced by Peng et al to consolidate the hetero-geneous networks into a single network architecture with 5Geven though they do not operate in the same bands to gain highspectral and operating and energy efficiency [107] Well known

Cloud Radio Access Network (C-RAN) architecture can per-form cooperative transmission across multiple edge nodes withcentralized cloud computing servers via fronthaul links [108]Although C-RAN provides high spectral efficiencies due tothe enhanced interference management capabilities with thecentralized baseband processing at the cloud it has potentiallylarge latencies F-RAN is proposed for 5G MEC deploymentsas an advanced socially aware mobile networking architectureto provide high spectral efficiency while maintaining highenergy efficiency and low latency [107] [108] Precodingdesign resource block allocation user scheduling and cellassociation are jointly designed for radio resource allocation inF-RANs in order to optimize spectral and energy efficienciesand latency performances [109] In [110] Rimal et al pro-pose a unified Time-Division Multiple Access (TDMA) basedresource management scheme for offloading traffic over Fiber-enabled Wireless (FiWi) access networks

In the envisioned 5G systems and MEC architecture bothbackhaul and wireless access links can be facilitated bymillimeter-Wave (mmW) spectrum [111] The use of mmWspectrum will enable high data rate access to MEC function-alities with low latency On the other hand MEC provideslocal computation power usefully for optimizing the perfor-mance of mmW communications In [112] [113] the authorsaddress the joint optimization of communicationcomputationresources with mmW communication They have taken the ad-vantage of blocking probabilities by considering intermittencyof mmW multi-link communications

An open source LPWAN infrastructure called OpenChirpis discussed in [114] OpenChirp which is developed usingLoRWAN allows multiple users to provision and to managebattery-powered transducers across large areas like campusesindustrial zones or cities As pointed out in [30] [115]SDN plays a vital role in improving MEC type technologiesby removing the technical shortcomings in edge computingimplementations The authors summarize the work performedfor implementing MEC based on NFV and SDN where theSDN controller manages the communication between MECservers which form a data center at the edge Table VIIIsummarizes the reviewed state-of-the-art communication is-sues and solutions in MEC enabled IoT

12

TABLE VIII Comparison of the reviewed state-of-the-art communication issues and solutions in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Commnetworkarchitecture

Commresourceallocation

[109] Performance analysis of radio resource allocation in F-RANs for edge cache andadaptive model selection to improve spectral efficiency and energy efficiency

Low latency andhigh reliability

X

[112][113]

Use of mmWave spectrum for high data rate access to MEC servers and backhaul links Low latency andhigh reliability

X

[114] An open source LPWAN infrastructure which allows multiple users to provision andmanage battery-powered transducers across large areas

LPWANinfrastructure

X

[42] A virtualized edge computing architecture with a proxy VM migration scheme tominimize traffic in the core network

IoT big datastreams

X

[115] Proposed network architecture includes multi-interface wireless access network(eg FiWi) heterogeneous backhauling distributed cloudlets hierarchical structure ofa cloudlet and the SDN based mobile core network

IoT big datastreams

X

[110] A novel unified resource management scheme for Ethernet-based FiWi networks thatjointly allocates bandwidth for transmissions of both conventional broadband traffic andMEC data in a TDMA fashion

Mission-criticalIoT

X

[116] Introduce Mobile-IoT-Federation-as-a-Service (MIFaaS) to enable dynamic cooperationamong privatepublic local clouds of IoT devices at the edge of the cellular infrastructureThe selection of the best configuration of federated IoT cloud platforms are modeledas a coalition formation problem

Cellular IoT X

[117] Allocation of radio resources in a joint LTE and NB-IoT system based of MIFaasparadigm [116] Discovered that in handling high-end IoT data traffic a combinationbetween NB-IoT and LTE is essential in providing the needed high data rate and lowlatency

Mission-criticalIoT

X

[118] Integration of D2D communications into edge computing environment reduce transmis-sion delay and traffic load across the network

Mission-criticalIoT

X

[119] Use the theories of stochastic geometry queueing and parallel computing for provi-sioning and planning MEC networks

Communicationlatency

X

C Computation Offloading and Resource Allocation

1) Requirements Computation offloading is the mostprominent and widely discussed feature of MEC that empow-ers resource-constrained IoT devices with augmented com-putational capabilities [29] [33] This will not only prolongthe battery life of the IoT sensor nodes but also reduceE2E latency needed to run sophisticated applications In thefirst place UE has to decide whether to execute the rela-tively simple tasks locally or offload to the MEC servers(ie task model for binary offloading) [32] Secondly thedecision of computation offloading to the MEC servers can beperformed fully or partially In the partial offloading a subsetof computations is executed locally while the rest is offloadedto the MEC server by considering several factors such asusers or application preferences (eg application buffer state)radio and backhaul connections quality (ie between UEand MEC servers) UE capabilities or cloud capabilities andavailability [29]

The sole objective of the offloading policies need to be theminimization of execution delay Other critical concerns areto define the dependency of offloadable components of theapplications based on their ability to partition data (eg real-time user input has to be processed at UE without offloading)and to predict the execution time of multiple tasks Theexecution order or routines have to be carefully formulatedsince certain outcomes can be the inputs of other tasks Aspointed out in [32] the task models for partial offloading canbe represented by task-call graphs with sequential paralleland general dependencies

Although in MEC computation offloading enables power-ful cloud services at the edge level the insufficient batteryenergy at the tiny IoT devices may incur new challenges In

applications like IoT surveillance or remote asset managementthe nodes are typically hard to reach Those applications mayalso require to offload data more frequently in small chunksby consuming more energy Therefore it is necessary to con-sider not only the trade-off between energy consumption andexecution delay in both full and partial offloading scenarios inMEC but also the trade-off between computation energy andtransmission energy consumption in order to extend batterylife

The joint computation and communication resource alloca-tion should be properly addressed in order to get the maximumutilization of available resources Single MEC server will beallocated for the applications which cannot be partitionedThe resources in multiple MEC servers are allocated for theoffloaded applications that can be split into several parts Whena job arrives at the MEC server if there are enough resourcesthe scheduler has to allocate the VM for further processingIf there are no sufficient computation resources it delegatesthe task to the centralized cloud MEC servers also have toallocate computation and communication resources for userapplication jobs and MEC service jobs User mobility networktopology network scalability and load balancing are someother factors to be considered in order to define fare resourceutilization policies on MEC servers Specifically when IoTgateways share limited bandwidth among multiple IoT deviceswhich can handle video audio or bio-medical signals theallocation of bandwidth will become challenging [120] Thelow power wireless technologies (eg BLE ZigBee lowpower Wi-Fi and LPWAN standards like LoRA or SigFox)used in IoT networks have limited bandwidth When the IoTdevices access the MEC server which is acting as the IoTgateway they have to utilize either of those low-power wireless

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

7

tructure vehicle to device vehicle to pedestrian vehicle tohome and vehicle to grid [54] In the context of IoT Au-tomotive V2X requires critical communication infrastructurewhere reliability and ultra low latency are crucial factors [55]Use cases in these categories include autonomous and semi-autonomous driving vehicle maintenance and in vehicle info-tainment In order to operate an efficient and reliable vehicularnetwork several features have to be improved these includereal-time traffic monitoring [56] [57] continuous sensing invehicles [58] [59] support for Infotainment applications [60]and improved security [61] However these features cannot beserved by current mobile networks [62] In this vein upcoming5G mobile systems are expected to offer a higher level of flexi-bility leveraging the emerging technologies related to networksoftwarization [63] In this context V2X combined with MECprovides a viable and cost-effective solution that can acceleratedevelopment of V2X and IoT automotive systems [64]

It is important to improve the performance of RAN tech-nologies to enable IoT automatization MEC will play a vitalrole here also For instance MEC technologies may fulfillthe latency reliability and throughput requirements in V2Xchannel modeling of mmWave communication [65] Moreoverthe placement of the MEC server within the RAN providesflexible network services for the vehicle and to efficientlycontrol the radio network resources [66] It is also possibleto design a time-predicted handover mechanism for vehiclesby leveraging road side information at MEC server in order tomeet the demand for high mobility and reliability in vehicularnetworks [66]

In addition ICN-MEC integration can also tackle existingtechnical challenges such as massive mobility of vehicles scal-ability deployment strategies service orchestration massivedata handling fast big data processing as well as ensuringsecurity and privacy [67]

Unmanned aerial vehicles (UAVs) or drones are anothertype of autonomous vehicles which are capable of sensingits environment and navigating without human inputs UAVuse cases include but not limited to public safety smart agri-culture surveillance and environmental monitoring [68] Inorder to maximize the flight time the UAV battery life shouldbe essentially conserved by minimizing the overhead onboardWhen the required processing power exceeds the availableresources on UAV the application data can be offloaded toMEC Accompanying the advanced RATs MEC will facilitatethe offloading process from UAV due to its expected widedeployment in the network [68]

D Gaming AR and VR

Mixed reality (MR) combines virtual reality (VR) andaugmented reality (AR) technologies thereby enabling humansto interact more naturally with the virtual worlds based ondata aggregated by IoT devices [69] With IoT AR tech-nologies are able to benefit directly from the high end inter-connection of objects that characterizes the IoT environmentthrough which users can extend their interactions from thereal world to the virtual world [2] [70] Convergence ofVR and IoT can occur in many ways such as telepresence

tourism industry smart transportation networks and roboticassisted surgeries Exclusive AR and VR experiences withthe delivery of 360 navigable videos will be offered byenhanced mobile broadband connections with low latencyand high reliability for mission-critical services With present-day network standards this might be impossible to achievehowever with the predicted characteristics of 5G such as20 Gbps peak data rate and 1 ms round-trip over-the-airlatency this becomes more easily achievable As identified byETSI MEC will be an ideal solution for low-latency offloadservices in AR and VR applications that combine computergenerated data with physical reality [71] While operatingVR devices over wireless links and deploying the VR controlcenter at MEC server the tracking accuracy can be increasedwith round trip latency of 1 ms and high reliability [72]Migrating computationally intensive tasks to edge servers willincrease the computational capacity of VR devices and savetheir battery-life Furthermore MEC will allow VR devices toaccess cloud resources in an on-demand fashion [73]

MEC platforms provide high capacity and low latencywireless coverage for large venues like stadiums or smartcities with a massive density of users to enjoy the AR andVR experience For instance inside a smart building with anetwork of cameras obtaining raw video frames and preparingthe processed frames for display can be performed locallywith the help of edge computing Furthermore tracking thelocal position of the user or object building a model of theenvironment and identifying known objects in the environ-ment can be offloaded to the edge cloud Similarly in orderto get absolute experience of VR glasses the response timeshould be extremely low When the user moves his head hemay experience delay if the glasses need to access remotedata centers Therefore the expected interaction time betweenmachines and humans needs to be less than 1ms When thelatency of a VR application is more than 1ms the user willexperience cyber sickness which will be interrupting the realVR experience MEC servers in the nearest proximity will beable to serve such applications with ultra low latency Futuregames will be played beyond the entertainment purposes ontop of VR and AR applications which would require theminimum possible latency Pokmon Go and Ingress are twoexamples of successful games that combine AR and sensorinformation such as user location

E Retail

The second largest MEC use case is expected to be in the re-tail businesses [37] Currently IoT has dominated retail marketapplications in many ways including digital signage supplychain management intelligent payment solutions smart vend-ing machines shelves doors resource management stream-ing and safety The high class retail stores which use facialrecognition systems need high definition cameras that generatehuge volumes of data requiring powerful servers within thepremises Therefore the on-site MEC servers will assist toprocess these kind of large data sets produced by IoT devicesin a retail market Big data analytics in shopping centerscan further exploit the collaborative processing between edge

8

and cloud computing [52] Installation of MEC in a retailmarket also provides high speed mobile coverage throughoutthe store WiFi access points that are maintained per store canbe connected to the MEC server to provide WiFi connectivityfor store customers as needed The enabling of MEC willalso omit load balancing Wi-Fi controllers or policy enginesrequired in the wide area networks in the store Althoughnot many academic published research works are explicitlyfocusing on MEC and IoT [74] they have become enormouslyreputed and commercialized technologies in the industry andthe business sectors

F Wearable IoT (WIoT)

During the previous years wearable technology has evolvedtremendously from walkman to step trackers smart watchesto smart glasses The development of low power wirelesstechnologies such as BLE (Bluetooth Low Energy) fuels thedevelopment of wearable devices Present-day wearables spanfrom low-end devices such as health and fitness trackers tohigh-end devices such as VRAR helmets and smart watchesIt is expected that wearables will become the worlds best-selling consumer electronics product after smartphones witha global availability of more than 929 million devices by2021 [75] With the new application domains and enablingservices wearable devices will demand more sophisticatedcommunication infrastructures For instance VRAR wear-ables are demanding gigabits throughput network connectivityto run their applications On the other hand dense deploymentof wearable devices in smart cities will increase the networktraffic on communication networks Thus the next generationcommunication networks should be able to provide the gigabitexperience for the anticipated ultra dense wearable devices[76]

Although cloud computing has enabled wide range of newnetworking services it cannot alone fulfill the upcoming re-quirements for the future wearable ecosystem Mainly the cen-tralized cloud data centers fails due to long End-to-End (E2E)latency Delay-sensitive wearable applications such as VRperceptual stability requires ultra low delay In this contextMEC has the potential to solve the limitations in current cloudbased systems by combining cloud and MEC infrastructuresThis will enable providers deploy storage computing andcaching capabilities in close proximity with such wearabledevices [76]

G IoT in Mechanized Agriculture

In order to meet the demands for future food productionthe agricultural sector will require some major evolution whereIoT will be integrated in various production management andanalytical processes [77] [78] The present-day agriculturalsector has been slow to adopting the emerging Machine-to-Machine (M2M) and IoT technologies when compared withother sectors like smart cities and the medical fields [79]

Precision farming and smart agriculture can be achievedusing autonomous vehicles (tractors) remote monitoring andreal-time analytics It is reported that farmers are increasinglyturning to agricultural drones and satellites to survey their

lands and generate crop data IoT sensors may provide in-formation about crop yields rainfall pest infestation and soilnutrition which are invaluable to production and can improvefarming techniques over time Although low latency is not acritical requirement in smart farming environment manage-ment of large data sets will be a key requirement to considerMEC servers located on-site can assist high tech farmingby collecting and analyzing big data on agriculture in orderto maximize efficiency Likewise without moving everydayfarming applications to a remote cloud MEC platforms canbenefit in terms of data access synchronization storage andother overhead costs the farmer might normally incur

The use of IoT-based automated data collection and moni-toring systems in poultry houses can be used to increase workefficiency and service quality and get a deeper understandingof chicken nurturing [80] Sensing technologies can be usedin carbon dioxide and luminosity sensing these are importantparameters in large scale poultry houses Gas sensors canbe used to get all necessary information to prevent chickeninfertility due to problems such as low carbon dioxide levelsLuminosity senors can help to maintain the proper luminositylevel for optimum productivity Similar to smart farms lowlatency is not a critical requirement in smart poultry houses[80] However it is critical to manage large data sets whereon-site MEC servers can be used In addition sharing the databetween poultry houses and storing legacy data in centralizedservers are important in identifying abnormal incidents in thefarm [81] With the use of MEC poultry houses can workwith intermittent connectivity to the centralized clouds In thatcase MEC servers can temporarily hold the data until farmsare connected with the centralized clouds

H Smart Energy

The smart grid system is an Information CommunicationTechnology (ICT)-enabled energy generation transmissionand distribution network It has capabilities to continuouslysense analyze and monitor both energy flow and energytransportation infrastructure Such features are enabled byadding digital controls and enabling network monitoring andtelecommunication capabilities As a result a smart grid doesnot only provide two-way flows of electrical power but alsoenables real-time automated bidirectional flow of informa-tion Adding such smartness to the aging energy infrastructurewill foster a more efficient energy system

IoT is considered as the foundation for realizing intelli-gence capabilities in smart grid systems IoT integrates theInternet-connectivity into all kinds of grid components suchas transformers breakers switches meters relays intelligentelectronic devices capacitor banks voltage regulators cam-eras and many more These IoT devices are then used tocapture the data required to enable automations IoT-enabledsmart grids provide several benefits such as reduced capitalexpenditure optimized renewable capacity lowered mainte-nance costs and enhanced customer engagement On one handthe transformation of an electrical grid into a smart systemrequires nearly every device and piece of equipment to havebuilt-in secure interconnected intelligence On the other hand

9

an efficient system is required to manage the generated dataie transferring storing and analyzing such huge amounts ofdata which are collected from these smart devices Thereforecloud computing is a viable solution to these IoT-based smartgrids [90]

Generally smart grids are spanning over large geograph-ical areas They often confront bandwidth bottlenecks andcommunication delays due to poor network connectivity andvast number of devices generating data Thus the traditionalcentralized cloud architecture is not suitable for the domain ofthe smart grid since it relies heavily on centralized processing[91] Many delay sensitive smart grid applications such asfault detection isolation and service restoration or VoltVARoptimization cannot tolerate round trip delay to access cen-tralized cloud systems MEC is identified as the viable cloud

computing option to address these limitations MEC allowsthe computation to be performed closer to the data sourceMoreover the potential attack points for the grid is increasingwith the growth of ubiquitous sensor deployment Everysmart IoT device can be vulnerable to potential attacks MECprovides the opportunity to enforce security mechanism closerto the end devices As such even if an attacker gains accessto an endpoint device the attack gets no further informationbeyond the local network segment since MEC has capabilitiesto notice the intrusion and cease the accessibility [85]

I Industrial Internet

The Industrial Internet of Things (IIoT) also known asIndustry 40 [92] is an application of IoT in the domain ofmanufacturing IIoT incorporates numerous advanced commu-

TABLE VI The reviewed state-of-the-art MEC integration in different IoT applications

Ref Description Smar

tH

ome

Smar

tC

ity

Hea

lthca

re

IoT

Aut

omot

ive

Gam

ing

AR

VR

Ret

ail

Wea

rabl

eIo

T

Smar

tA

gric

ultu

re

Smar

tE

nerg

y

Indu

stri

alIn

tern

et

[40] Preliminary design of deploying MEC server functionalities in a smart hometo realize IoT gateway with direct M2M interaction in LTE networks

X

[41] Introduce Gateway-as-a-Service for heterogeneous IoT devices on top of thevirtualization technologies in edge computing

X X

[44] Propose an autonomic creation of MEC services to enhance QoS of videostreaming in smart cities

X

[50] Propose a semantic edge-based IoT architecture for military health services inbattlefield

X

[51] Provide a conceptual MEC based architecture for mission-critical context awarecollaboration in remote surgeries

X

[53] Describe and implement a smart e-Health gateway at the edge of the networksuitable for ubiquitous healthcare systems

X

[64] Analysis on research and engineering challenges co-existence of cloud edgecomputing and data caching strategies at the edge for vehicular networks

X

[82] Discuss the design aspects for the radio access in 5G V2X X[65] Discuss the benefits of merging MEC and mmWave technologies for 5G

applicationsX X X

[66] Propose a novel MEC-based architecture for future cellular vehicular networks X[67] Discuss the benefits of combining ICN and MEC in the context of connected

vehicle environmentsX

[52] Propose a framework for big data analytics between edge and cloud computingplatforms

X X

[74] Design and implement a fog computing based framework that support sharingand reusing contextual data across services in smart city and retail stores

X X

[83] Present a usecase of MEC for Tactile Internet based 5G gaming application X[84] A demonstration of MEC for Tactile Internet based 5G gaming application X[76] Discuss the role of MEC in 5G WIoT communication and its challenges X X[68] Propose an UAV-based IoT platform for a crowd surveillance use case X X[78] Develop and test a ubiquitous sensor network platform for crop lands automa-

tion maintenance in precision agricultureX

[70] Present a serverless edge computing architecture that enables the offloading ofmobile computation with low latency and high throughput using a mobile ARapplication

X

[85] Discuss the benefits of MEC and edge computing (EC) to enhance the securityof smart grids

X

[86] Present a method to optimize the EC based video streaming schemes forIndustrial IoT

X

[87] Present the use of edge computing to provide elastic resources and services toenable microdatabases architecture for IIoT

X

[88] Propose a fog-based communication architecture for Industry 40 applications X[73] Describe research directions and enablers of wireless interconnected VR

systemsX

[89] Design an optimization framework for VRAR communication via small-cellcooperation

X

10

nication and automation technologies such as M2M commu-nication machine learning and big data analytics to improveintelligence and the connectivity [93] For instance IIoTnetworks can connect all of the employees data and processesfrom the factory floor and forward them to the executiveoffices Thus decision makers or employees can create a fulland accurate view of their manufacturing process by usingIIoT network hence improving their ability to make moreinformed decisions IIoT also helps the exploitation as well asimplementation of new intelligent technologies to acceleratethe innovation and transformation of the factory workforce[92]

Primarily IIoT is seen as a way to improve operationalefficiency However IIoT provides a wide range of otherbenefits such as improving connectivity efficiency scalabilitytime savings as well as cost savings for manufacturing pro-cesses with the maximum use of smart machines [92] [94] Ingeneral these smart machines operate with higher accuracygreater efficiency and constant working capabilities than hu-mans [95] Thus IIoT has great potential for improving qualitycontrol sustainability and overall supply chain efficiency

MEC will play a vital role in enabling future IIoT applica-tions [96] by addressing the shortcomings of M2M communi-cation (eg latency resilience cost peer-to-peer connectivitysecurity) in IIoT domain [97] [98] Current market trendsalready show that edge computing will represent many im-plementation scenarios for IIoT For instance real-time edgeanalytics and enhanced edge security are two key drivers inthe creation of new IIoT deployments Thus the addition ofMEC in IIoT networks will fuel the evolution of IIoT as wellas create new business applications [99]

One way to optimize the use of conventional edge com-puting in video streaming schemes for IIoT is presented in[86] By using machine learning algorithms edge computingcan process the sensor data before transmitting to the cloudThis mitigates against the degradation of service quality of thevideo streaming Aggregation of all the sensor data to a singledata center increases latency and raises performance concernsin IIoT domain In order to solve this issue a microdatabasearchitecture is proposed for the Industrial Internet [87] It holdsthe data close to the industrial processes but also makes itavailable near the applications that can benefit from the dataEdge computing also provides elastic resources and services toenable micro-database architecture [87] A fog-based commu-nication architecture for Industry 40 applications is proposedin [88] This approach will substantially minimize the energyconsumption of the IoT nodes Edge computational capabilitiesare further used to predict future data measurements andreduce the throughput from IoT devices to the control unit

III TECHNICAL ASPECTS OF MEC ENABLED IOT

To realize the MEC exploitation for IoT applications thekey value propositions are mostly seen from the technicalparameters such as scalability communication computationoffloading and resource allocation mobility management se-curity privacy and trust management This section describesthe state-of-the-art of each of these technical parameters hence

giving a clear background against which the benefits of MECcan be envisioned

A Scalability

1) Requirements When it comes to actual deployment ofMEC platform for IoT systems scalability is a key factorto consider The compatibility of MEC servers to multiplenetwork environments is one of the factors that will driveits large scale adoption in future networks [100] The IoTenvironment will consist of hundreds of billions of sensors ac-tuators Radio-Frequency Identification (RFID)-tagged objectssoftware vehicles and embedded systems all interconnectedin a huge network of cyber-physical systems At a utility scaleconsideration these devices will be working in close collab-oration to deliver the expected services in technologies likethe smart grids virtual power plants smart homes intelligenttransportation and smart cities That being said the role ofscalability to the realization of such a hyper-connected IoTenvironment becomes more obvious The IoT environment willrequire a dynamic range of capabilities in the network space ifsuch large numbers of devices are to be supported effectively

2) Related work Currently MEC servers have been con-firmed to be compatible with LTE macro base station (eN-odeB) sites 3G Radio Network Controller (RNC) site multi-Radio Access Technology (RAT) cell aggregation site andat the edge of the core network [2] Such multi-RAT cellaggregation schemes can be implemented indoor or outdoorsettings depending on the requirements This invariably en-ables MEC to be applied to many different possible scenariosThe larger the deployment scenarios for MEC the more therange of capabilities it can handle this also translates to higherscalability for MEC-enable technologies like IoT

Designing an edge cloud network implies that an optimallocation for citing the cloud facility is first determined In[105] authors present a design optimization scheme for theMEC architecture based on link-path formulation supportedby heuristics in order to optimize the computation time forthe scheme In this approach consideration is given to bothusers and VMs mobility Hence an optimal point to installthe MEC server is determined through a tread-off betweeninstallation cost and the quality of service to be deliveredTable VII compares the reviewed state-of-the-art scalabilityfeature in MEC enabled IoT

B Communication

1) Requirements There are three main categories for thecommunication concerns about MEC [106] Wireless accesswhile offloading to the mobile edge host Backhaul accesswhile offloading to a remote cloud server Communicationamong IoT devices mobile edge host and remote cloudservers when they collaboratively execute multiple jobs Thefirst and the second categories are the most renowned on behalfof the MEC servers which are the small scale data centersdeployed by the network operators and can be co-located withthe Wireless Access Points (WAPs) In the IoT supportiveMEC systems the consumer devices may communicate with

11

TABLE VII Comparison of the reviewed state-of-the-art scalability feature in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Addressing Search

[101] Discusses the challenges in searching imposed by the burgeoning field of IoT General IoT X[94] Examines a variety of popular and innovative IoT solutions in terms of context-aware

technology perspectives to serve as a conceptual framework for context-aware productdevelopment and research in the IoT paradigm

General IoT solu-tions

X

[102] Proposes an innovative distributed architecture combining machine-to-machine industry-mature protocols (ie MQTT and CoAP) in an original way to enhance the scalabilityof gateways for the efficient IoT-cloud integration

IoT cloud inte-gration

X X

[103] Studies an implementation of edge computing which exploits transparent computing tobuild scalable IoT platforms using transparent computing

Wearable IoT X

[104] Introduces a lightweight edge gateway for the IoT architecture using container-basedvirtualization techniques

General IoT X

the MEC servers either directly or with the support of neigh-boring devices using Device-to-Device (D2D) communicationFor the third category WAPs enable access to the remote datacenters in the central cloud through backhaul links

In order to reap the maximum advantage of computationoffloading leveraged at the edge servers MEC systems needefficient communication channels Unlike the wired connec-tions in the conventional grid computing and cloud computingthe wireless access links between the mobile devices andcloud computing resources in the edge computing paradigmcan be unstable Sudden service outages may occur with theinterruption of access links The inherent challenges withwireless communication channels like multi-path fading in-terference and spectrum shortage should always be taken intoaccount for the design of MEC systems to seamlessly integratecomputation offloading and radio resource management [32]Moreover both wireless and backhaul access links have lim-ited capacities which should be properly shared among mobiledevices in a similar way as sharing the computing resources ofthe MEC server Hence having a cooperative scheme for thejoint allocation of communication and computation resourcesis important for the successful deployment of MEC [106]Redesigning both communication and networking protocolsto integrate communication infrastructures in MEC and IoTsystems is a challenging task The key focus should be onimproving the computation efficiency with respect to datatransmission

Another major requirement is to maintain interoperabilitywhile addressing heterogeneous communication technologiesthat have to be utilized in IoT and MEC paradigms in 5GThere are plenty of radio technologies that facilitate IoT Low-Power Wide Area Networks (LPWANs) (eg WCDMA LTEnarrowband IoT (NB-IoT) Wi-Fi Bluetooth Zigbee SIGFOXand LoRA) The choice of these LPWAN technologies maycreate trade-offs among signal strength operational rangethroughput and power consumption With the arrival of 5Gthe convergence of these communication technologies needsto be achieved since one network will not be fitting based onthose trade-offs

2) Related work Recently Fog-Radio Access Network (F-RAN) was introduced by Peng et al to consolidate the hetero-geneous networks into a single network architecture with 5Geven though they do not operate in the same bands to gain highspectral and operating and energy efficiency [107] Well known

Cloud Radio Access Network (C-RAN) architecture can per-form cooperative transmission across multiple edge nodes withcentralized cloud computing servers via fronthaul links [108]Although C-RAN provides high spectral efficiencies due tothe enhanced interference management capabilities with thecentralized baseband processing at the cloud it has potentiallylarge latencies F-RAN is proposed for 5G MEC deploymentsas an advanced socially aware mobile networking architectureto provide high spectral efficiency while maintaining highenergy efficiency and low latency [107] [108] Precodingdesign resource block allocation user scheduling and cellassociation are jointly designed for radio resource allocation inF-RANs in order to optimize spectral and energy efficienciesand latency performances [109] In [110] Rimal et al pro-pose a unified Time-Division Multiple Access (TDMA) basedresource management scheme for offloading traffic over Fiber-enabled Wireless (FiWi) access networks

In the envisioned 5G systems and MEC architecture bothbackhaul and wireless access links can be facilitated bymillimeter-Wave (mmW) spectrum [111] The use of mmWspectrum will enable high data rate access to MEC function-alities with low latency On the other hand MEC provideslocal computation power usefully for optimizing the perfor-mance of mmW communications In [112] [113] the authorsaddress the joint optimization of communicationcomputationresources with mmW communication They have taken the ad-vantage of blocking probabilities by considering intermittencyof mmW multi-link communications

An open source LPWAN infrastructure called OpenChirpis discussed in [114] OpenChirp which is developed usingLoRWAN allows multiple users to provision and to managebattery-powered transducers across large areas like campusesindustrial zones or cities As pointed out in [30] [115]SDN plays a vital role in improving MEC type technologiesby removing the technical shortcomings in edge computingimplementations The authors summarize the work performedfor implementing MEC based on NFV and SDN where theSDN controller manages the communication between MECservers which form a data center at the edge Table VIIIsummarizes the reviewed state-of-the-art communication is-sues and solutions in MEC enabled IoT

12

TABLE VIII Comparison of the reviewed state-of-the-art communication issues and solutions in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Commnetworkarchitecture

Commresourceallocation

[109] Performance analysis of radio resource allocation in F-RANs for edge cache andadaptive model selection to improve spectral efficiency and energy efficiency

Low latency andhigh reliability

X

[112][113]

Use of mmWave spectrum for high data rate access to MEC servers and backhaul links Low latency andhigh reliability

X

[114] An open source LPWAN infrastructure which allows multiple users to provision andmanage battery-powered transducers across large areas

LPWANinfrastructure

X

[42] A virtualized edge computing architecture with a proxy VM migration scheme tominimize traffic in the core network

IoT big datastreams

X

[115] Proposed network architecture includes multi-interface wireless access network(eg FiWi) heterogeneous backhauling distributed cloudlets hierarchical structure ofa cloudlet and the SDN based mobile core network

IoT big datastreams

X

[110] A novel unified resource management scheme for Ethernet-based FiWi networks thatjointly allocates bandwidth for transmissions of both conventional broadband traffic andMEC data in a TDMA fashion

Mission-criticalIoT

X

[116] Introduce Mobile-IoT-Federation-as-a-Service (MIFaaS) to enable dynamic cooperationamong privatepublic local clouds of IoT devices at the edge of the cellular infrastructureThe selection of the best configuration of federated IoT cloud platforms are modeledas a coalition formation problem

Cellular IoT X

[117] Allocation of radio resources in a joint LTE and NB-IoT system based of MIFaasparadigm [116] Discovered that in handling high-end IoT data traffic a combinationbetween NB-IoT and LTE is essential in providing the needed high data rate and lowlatency

Mission-criticalIoT

X

[118] Integration of D2D communications into edge computing environment reduce transmis-sion delay and traffic load across the network

Mission-criticalIoT

X

[119] Use the theories of stochastic geometry queueing and parallel computing for provi-sioning and planning MEC networks

Communicationlatency

X

C Computation Offloading and Resource Allocation

1) Requirements Computation offloading is the mostprominent and widely discussed feature of MEC that empow-ers resource-constrained IoT devices with augmented com-putational capabilities [29] [33] This will not only prolongthe battery life of the IoT sensor nodes but also reduceE2E latency needed to run sophisticated applications In thefirst place UE has to decide whether to execute the rela-tively simple tasks locally or offload to the MEC servers(ie task model for binary offloading) [32] Secondly thedecision of computation offloading to the MEC servers can beperformed fully or partially In the partial offloading a subsetof computations is executed locally while the rest is offloadedto the MEC server by considering several factors such asusers or application preferences (eg application buffer state)radio and backhaul connections quality (ie between UEand MEC servers) UE capabilities or cloud capabilities andavailability [29]

The sole objective of the offloading policies need to be theminimization of execution delay Other critical concerns areto define the dependency of offloadable components of theapplications based on their ability to partition data (eg real-time user input has to be processed at UE without offloading)and to predict the execution time of multiple tasks Theexecution order or routines have to be carefully formulatedsince certain outcomes can be the inputs of other tasks Aspointed out in [32] the task models for partial offloading canbe represented by task-call graphs with sequential paralleland general dependencies

Although in MEC computation offloading enables power-ful cloud services at the edge level the insufficient batteryenergy at the tiny IoT devices may incur new challenges In

applications like IoT surveillance or remote asset managementthe nodes are typically hard to reach Those applications mayalso require to offload data more frequently in small chunksby consuming more energy Therefore it is necessary to con-sider not only the trade-off between energy consumption andexecution delay in both full and partial offloading scenarios inMEC but also the trade-off between computation energy andtransmission energy consumption in order to extend batterylife

The joint computation and communication resource alloca-tion should be properly addressed in order to get the maximumutilization of available resources Single MEC server will beallocated for the applications which cannot be partitionedThe resources in multiple MEC servers are allocated for theoffloaded applications that can be split into several parts Whena job arrives at the MEC server if there are enough resourcesthe scheduler has to allocate the VM for further processingIf there are no sufficient computation resources it delegatesthe task to the centralized cloud MEC servers also have toallocate computation and communication resources for userapplication jobs and MEC service jobs User mobility networktopology network scalability and load balancing are someother factors to be considered in order to define fare resourceutilization policies on MEC servers Specifically when IoTgateways share limited bandwidth among multiple IoT deviceswhich can handle video audio or bio-medical signals theallocation of bandwidth will become challenging [120] Thelow power wireless technologies (eg BLE ZigBee lowpower Wi-Fi and LPWAN standards like LoRA or SigFox)used in IoT networks have limited bandwidth When the IoTdevices access the MEC server which is acting as the IoTgateway they have to utilize either of those low-power wireless

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

8

and cloud computing [52] Installation of MEC in a retailmarket also provides high speed mobile coverage throughoutthe store WiFi access points that are maintained per store canbe connected to the MEC server to provide WiFi connectivityfor store customers as needed The enabling of MEC willalso omit load balancing Wi-Fi controllers or policy enginesrequired in the wide area networks in the store Althoughnot many academic published research works are explicitlyfocusing on MEC and IoT [74] they have become enormouslyreputed and commercialized technologies in the industry andthe business sectors

F Wearable IoT (WIoT)

During the previous years wearable technology has evolvedtremendously from walkman to step trackers smart watchesto smart glasses The development of low power wirelesstechnologies such as BLE (Bluetooth Low Energy) fuels thedevelopment of wearable devices Present-day wearables spanfrom low-end devices such as health and fitness trackers tohigh-end devices such as VRAR helmets and smart watchesIt is expected that wearables will become the worlds best-selling consumer electronics product after smartphones witha global availability of more than 929 million devices by2021 [75] With the new application domains and enablingservices wearable devices will demand more sophisticatedcommunication infrastructures For instance VRAR wear-ables are demanding gigabits throughput network connectivityto run their applications On the other hand dense deploymentof wearable devices in smart cities will increase the networktraffic on communication networks Thus the next generationcommunication networks should be able to provide the gigabitexperience for the anticipated ultra dense wearable devices[76]

Although cloud computing has enabled wide range of newnetworking services it cannot alone fulfill the upcoming re-quirements for the future wearable ecosystem Mainly the cen-tralized cloud data centers fails due to long End-to-End (E2E)latency Delay-sensitive wearable applications such as VRperceptual stability requires ultra low delay In this contextMEC has the potential to solve the limitations in current cloudbased systems by combining cloud and MEC infrastructuresThis will enable providers deploy storage computing andcaching capabilities in close proximity with such wearabledevices [76]

G IoT in Mechanized Agriculture

In order to meet the demands for future food productionthe agricultural sector will require some major evolution whereIoT will be integrated in various production management andanalytical processes [77] [78] The present-day agriculturalsector has been slow to adopting the emerging Machine-to-Machine (M2M) and IoT technologies when compared withother sectors like smart cities and the medical fields [79]

Precision farming and smart agriculture can be achievedusing autonomous vehicles (tractors) remote monitoring andreal-time analytics It is reported that farmers are increasinglyturning to agricultural drones and satellites to survey their

lands and generate crop data IoT sensors may provide in-formation about crop yields rainfall pest infestation and soilnutrition which are invaluable to production and can improvefarming techniques over time Although low latency is not acritical requirement in smart farming environment manage-ment of large data sets will be a key requirement to considerMEC servers located on-site can assist high tech farmingby collecting and analyzing big data on agriculture in orderto maximize efficiency Likewise without moving everydayfarming applications to a remote cloud MEC platforms canbenefit in terms of data access synchronization storage andother overhead costs the farmer might normally incur

The use of IoT-based automated data collection and moni-toring systems in poultry houses can be used to increase workefficiency and service quality and get a deeper understandingof chicken nurturing [80] Sensing technologies can be usedin carbon dioxide and luminosity sensing these are importantparameters in large scale poultry houses Gas sensors canbe used to get all necessary information to prevent chickeninfertility due to problems such as low carbon dioxide levelsLuminosity senors can help to maintain the proper luminositylevel for optimum productivity Similar to smart farms lowlatency is not a critical requirement in smart poultry houses[80] However it is critical to manage large data sets whereon-site MEC servers can be used In addition sharing the databetween poultry houses and storing legacy data in centralizedservers are important in identifying abnormal incidents in thefarm [81] With the use of MEC poultry houses can workwith intermittent connectivity to the centralized clouds In thatcase MEC servers can temporarily hold the data until farmsare connected with the centralized clouds

H Smart Energy

The smart grid system is an Information CommunicationTechnology (ICT)-enabled energy generation transmissionand distribution network It has capabilities to continuouslysense analyze and monitor both energy flow and energytransportation infrastructure Such features are enabled byadding digital controls and enabling network monitoring andtelecommunication capabilities As a result a smart grid doesnot only provide two-way flows of electrical power but alsoenables real-time automated bidirectional flow of informa-tion Adding such smartness to the aging energy infrastructurewill foster a more efficient energy system

IoT is considered as the foundation for realizing intelli-gence capabilities in smart grid systems IoT integrates theInternet-connectivity into all kinds of grid components suchas transformers breakers switches meters relays intelligentelectronic devices capacitor banks voltage regulators cam-eras and many more These IoT devices are then used tocapture the data required to enable automations IoT-enabledsmart grids provide several benefits such as reduced capitalexpenditure optimized renewable capacity lowered mainte-nance costs and enhanced customer engagement On one handthe transformation of an electrical grid into a smart systemrequires nearly every device and piece of equipment to havebuilt-in secure interconnected intelligence On the other hand

9

an efficient system is required to manage the generated dataie transferring storing and analyzing such huge amounts ofdata which are collected from these smart devices Thereforecloud computing is a viable solution to these IoT-based smartgrids [90]

Generally smart grids are spanning over large geograph-ical areas They often confront bandwidth bottlenecks andcommunication delays due to poor network connectivity andvast number of devices generating data Thus the traditionalcentralized cloud architecture is not suitable for the domain ofthe smart grid since it relies heavily on centralized processing[91] Many delay sensitive smart grid applications such asfault detection isolation and service restoration or VoltVARoptimization cannot tolerate round trip delay to access cen-tralized cloud systems MEC is identified as the viable cloud

computing option to address these limitations MEC allowsthe computation to be performed closer to the data sourceMoreover the potential attack points for the grid is increasingwith the growth of ubiquitous sensor deployment Everysmart IoT device can be vulnerable to potential attacks MECprovides the opportunity to enforce security mechanism closerto the end devices As such even if an attacker gains accessto an endpoint device the attack gets no further informationbeyond the local network segment since MEC has capabilitiesto notice the intrusion and cease the accessibility [85]

I Industrial Internet

The Industrial Internet of Things (IIoT) also known asIndustry 40 [92] is an application of IoT in the domain ofmanufacturing IIoT incorporates numerous advanced commu-

TABLE VI The reviewed state-of-the-art MEC integration in different IoT applications

Ref Description Smar

tH

ome

Smar

tC

ity

Hea

lthca

re

IoT

Aut

omot

ive

Gam

ing

AR

VR

Ret

ail

Wea

rabl

eIo

T

Smar

tA

gric

ultu

re

Smar

tE

nerg

y

Indu

stri

alIn

tern

et

[40] Preliminary design of deploying MEC server functionalities in a smart hometo realize IoT gateway with direct M2M interaction in LTE networks

X

[41] Introduce Gateway-as-a-Service for heterogeneous IoT devices on top of thevirtualization technologies in edge computing

X X

[44] Propose an autonomic creation of MEC services to enhance QoS of videostreaming in smart cities

X

[50] Propose a semantic edge-based IoT architecture for military health services inbattlefield

X

[51] Provide a conceptual MEC based architecture for mission-critical context awarecollaboration in remote surgeries

X

[53] Describe and implement a smart e-Health gateway at the edge of the networksuitable for ubiquitous healthcare systems

X

[64] Analysis on research and engineering challenges co-existence of cloud edgecomputing and data caching strategies at the edge for vehicular networks

X

[82] Discuss the design aspects for the radio access in 5G V2X X[65] Discuss the benefits of merging MEC and mmWave technologies for 5G

applicationsX X X

[66] Propose a novel MEC-based architecture for future cellular vehicular networks X[67] Discuss the benefits of combining ICN and MEC in the context of connected

vehicle environmentsX

[52] Propose a framework for big data analytics between edge and cloud computingplatforms

X X

[74] Design and implement a fog computing based framework that support sharingand reusing contextual data across services in smart city and retail stores

X X

[83] Present a usecase of MEC for Tactile Internet based 5G gaming application X[84] A demonstration of MEC for Tactile Internet based 5G gaming application X[76] Discuss the role of MEC in 5G WIoT communication and its challenges X X[68] Propose an UAV-based IoT platform for a crowd surveillance use case X X[78] Develop and test a ubiquitous sensor network platform for crop lands automa-

tion maintenance in precision agricultureX

[70] Present a serverless edge computing architecture that enables the offloading ofmobile computation with low latency and high throughput using a mobile ARapplication

X

[85] Discuss the benefits of MEC and edge computing (EC) to enhance the securityof smart grids

X

[86] Present a method to optimize the EC based video streaming schemes forIndustrial IoT

X

[87] Present the use of edge computing to provide elastic resources and services toenable microdatabases architecture for IIoT

X

[88] Propose a fog-based communication architecture for Industry 40 applications X[73] Describe research directions and enablers of wireless interconnected VR

systemsX

[89] Design an optimization framework for VRAR communication via small-cellcooperation

X

10

nication and automation technologies such as M2M commu-nication machine learning and big data analytics to improveintelligence and the connectivity [93] For instance IIoTnetworks can connect all of the employees data and processesfrom the factory floor and forward them to the executiveoffices Thus decision makers or employees can create a fulland accurate view of their manufacturing process by usingIIoT network hence improving their ability to make moreinformed decisions IIoT also helps the exploitation as well asimplementation of new intelligent technologies to acceleratethe innovation and transformation of the factory workforce[92]

Primarily IIoT is seen as a way to improve operationalefficiency However IIoT provides a wide range of otherbenefits such as improving connectivity efficiency scalabilitytime savings as well as cost savings for manufacturing pro-cesses with the maximum use of smart machines [92] [94] Ingeneral these smart machines operate with higher accuracygreater efficiency and constant working capabilities than hu-mans [95] Thus IIoT has great potential for improving qualitycontrol sustainability and overall supply chain efficiency

MEC will play a vital role in enabling future IIoT applica-tions [96] by addressing the shortcomings of M2M communi-cation (eg latency resilience cost peer-to-peer connectivitysecurity) in IIoT domain [97] [98] Current market trendsalready show that edge computing will represent many im-plementation scenarios for IIoT For instance real-time edgeanalytics and enhanced edge security are two key drivers inthe creation of new IIoT deployments Thus the addition ofMEC in IIoT networks will fuel the evolution of IIoT as wellas create new business applications [99]

One way to optimize the use of conventional edge com-puting in video streaming schemes for IIoT is presented in[86] By using machine learning algorithms edge computingcan process the sensor data before transmitting to the cloudThis mitigates against the degradation of service quality of thevideo streaming Aggregation of all the sensor data to a singledata center increases latency and raises performance concernsin IIoT domain In order to solve this issue a microdatabasearchitecture is proposed for the Industrial Internet [87] It holdsthe data close to the industrial processes but also makes itavailable near the applications that can benefit from the dataEdge computing also provides elastic resources and services toenable micro-database architecture [87] A fog-based commu-nication architecture for Industry 40 applications is proposedin [88] This approach will substantially minimize the energyconsumption of the IoT nodes Edge computational capabilitiesare further used to predict future data measurements andreduce the throughput from IoT devices to the control unit

III TECHNICAL ASPECTS OF MEC ENABLED IOT

To realize the MEC exploitation for IoT applications thekey value propositions are mostly seen from the technicalparameters such as scalability communication computationoffloading and resource allocation mobility management se-curity privacy and trust management This section describesthe state-of-the-art of each of these technical parameters hence

giving a clear background against which the benefits of MECcan be envisioned

A Scalability

1) Requirements When it comes to actual deployment ofMEC platform for IoT systems scalability is a key factorto consider The compatibility of MEC servers to multiplenetwork environments is one of the factors that will driveits large scale adoption in future networks [100] The IoTenvironment will consist of hundreds of billions of sensors ac-tuators Radio-Frequency Identification (RFID)-tagged objectssoftware vehicles and embedded systems all interconnectedin a huge network of cyber-physical systems At a utility scaleconsideration these devices will be working in close collab-oration to deliver the expected services in technologies likethe smart grids virtual power plants smart homes intelligenttransportation and smart cities That being said the role ofscalability to the realization of such a hyper-connected IoTenvironment becomes more obvious The IoT environment willrequire a dynamic range of capabilities in the network space ifsuch large numbers of devices are to be supported effectively

2) Related work Currently MEC servers have been con-firmed to be compatible with LTE macro base station (eN-odeB) sites 3G Radio Network Controller (RNC) site multi-Radio Access Technology (RAT) cell aggregation site andat the edge of the core network [2] Such multi-RAT cellaggregation schemes can be implemented indoor or outdoorsettings depending on the requirements This invariably en-ables MEC to be applied to many different possible scenariosThe larger the deployment scenarios for MEC the more therange of capabilities it can handle this also translates to higherscalability for MEC-enable technologies like IoT

Designing an edge cloud network implies that an optimallocation for citing the cloud facility is first determined In[105] authors present a design optimization scheme for theMEC architecture based on link-path formulation supportedby heuristics in order to optimize the computation time forthe scheme In this approach consideration is given to bothusers and VMs mobility Hence an optimal point to installthe MEC server is determined through a tread-off betweeninstallation cost and the quality of service to be deliveredTable VII compares the reviewed state-of-the-art scalabilityfeature in MEC enabled IoT

B Communication

1) Requirements There are three main categories for thecommunication concerns about MEC [106] Wireless accesswhile offloading to the mobile edge host Backhaul accesswhile offloading to a remote cloud server Communicationamong IoT devices mobile edge host and remote cloudservers when they collaboratively execute multiple jobs Thefirst and the second categories are the most renowned on behalfof the MEC servers which are the small scale data centersdeployed by the network operators and can be co-located withthe Wireless Access Points (WAPs) In the IoT supportiveMEC systems the consumer devices may communicate with

11

TABLE VII Comparison of the reviewed state-of-the-art scalability feature in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Addressing Search

[101] Discusses the challenges in searching imposed by the burgeoning field of IoT General IoT X[94] Examines a variety of popular and innovative IoT solutions in terms of context-aware

technology perspectives to serve as a conceptual framework for context-aware productdevelopment and research in the IoT paradigm

General IoT solu-tions

X

[102] Proposes an innovative distributed architecture combining machine-to-machine industry-mature protocols (ie MQTT and CoAP) in an original way to enhance the scalabilityof gateways for the efficient IoT-cloud integration

IoT cloud inte-gration

X X

[103] Studies an implementation of edge computing which exploits transparent computing tobuild scalable IoT platforms using transparent computing

Wearable IoT X

[104] Introduces a lightweight edge gateway for the IoT architecture using container-basedvirtualization techniques

General IoT X

the MEC servers either directly or with the support of neigh-boring devices using Device-to-Device (D2D) communicationFor the third category WAPs enable access to the remote datacenters in the central cloud through backhaul links

In order to reap the maximum advantage of computationoffloading leveraged at the edge servers MEC systems needefficient communication channels Unlike the wired connec-tions in the conventional grid computing and cloud computingthe wireless access links between the mobile devices andcloud computing resources in the edge computing paradigmcan be unstable Sudden service outages may occur with theinterruption of access links The inherent challenges withwireless communication channels like multi-path fading in-terference and spectrum shortage should always be taken intoaccount for the design of MEC systems to seamlessly integratecomputation offloading and radio resource management [32]Moreover both wireless and backhaul access links have lim-ited capacities which should be properly shared among mobiledevices in a similar way as sharing the computing resources ofthe MEC server Hence having a cooperative scheme for thejoint allocation of communication and computation resourcesis important for the successful deployment of MEC [106]Redesigning both communication and networking protocolsto integrate communication infrastructures in MEC and IoTsystems is a challenging task The key focus should be onimproving the computation efficiency with respect to datatransmission

Another major requirement is to maintain interoperabilitywhile addressing heterogeneous communication technologiesthat have to be utilized in IoT and MEC paradigms in 5GThere are plenty of radio technologies that facilitate IoT Low-Power Wide Area Networks (LPWANs) (eg WCDMA LTEnarrowband IoT (NB-IoT) Wi-Fi Bluetooth Zigbee SIGFOXand LoRA) The choice of these LPWAN technologies maycreate trade-offs among signal strength operational rangethroughput and power consumption With the arrival of 5Gthe convergence of these communication technologies needsto be achieved since one network will not be fitting based onthose trade-offs

2) Related work Recently Fog-Radio Access Network (F-RAN) was introduced by Peng et al to consolidate the hetero-geneous networks into a single network architecture with 5Geven though they do not operate in the same bands to gain highspectral and operating and energy efficiency [107] Well known

Cloud Radio Access Network (C-RAN) architecture can per-form cooperative transmission across multiple edge nodes withcentralized cloud computing servers via fronthaul links [108]Although C-RAN provides high spectral efficiencies due tothe enhanced interference management capabilities with thecentralized baseband processing at the cloud it has potentiallylarge latencies F-RAN is proposed for 5G MEC deploymentsas an advanced socially aware mobile networking architectureto provide high spectral efficiency while maintaining highenergy efficiency and low latency [107] [108] Precodingdesign resource block allocation user scheduling and cellassociation are jointly designed for radio resource allocation inF-RANs in order to optimize spectral and energy efficienciesand latency performances [109] In [110] Rimal et al pro-pose a unified Time-Division Multiple Access (TDMA) basedresource management scheme for offloading traffic over Fiber-enabled Wireless (FiWi) access networks

In the envisioned 5G systems and MEC architecture bothbackhaul and wireless access links can be facilitated bymillimeter-Wave (mmW) spectrum [111] The use of mmWspectrum will enable high data rate access to MEC function-alities with low latency On the other hand MEC provideslocal computation power usefully for optimizing the perfor-mance of mmW communications In [112] [113] the authorsaddress the joint optimization of communicationcomputationresources with mmW communication They have taken the ad-vantage of blocking probabilities by considering intermittencyof mmW multi-link communications

An open source LPWAN infrastructure called OpenChirpis discussed in [114] OpenChirp which is developed usingLoRWAN allows multiple users to provision and to managebattery-powered transducers across large areas like campusesindustrial zones or cities As pointed out in [30] [115]SDN plays a vital role in improving MEC type technologiesby removing the technical shortcomings in edge computingimplementations The authors summarize the work performedfor implementing MEC based on NFV and SDN where theSDN controller manages the communication between MECservers which form a data center at the edge Table VIIIsummarizes the reviewed state-of-the-art communication is-sues and solutions in MEC enabled IoT

12

TABLE VIII Comparison of the reviewed state-of-the-art communication issues and solutions in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Commnetworkarchitecture

Commresourceallocation

[109] Performance analysis of radio resource allocation in F-RANs for edge cache andadaptive model selection to improve spectral efficiency and energy efficiency

Low latency andhigh reliability

X

[112][113]

Use of mmWave spectrum for high data rate access to MEC servers and backhaul links Low latency andhigh reliability

X

[114] An open source LPWAN infrastructure which allows multiple users to provision andmanage battery-powered transducers across large areas

LPWANinfrastructure

X

[42] A virtualized edge computing architecture with a proxy VM migration scheme tominimize traffic in the core network

IoT big datastreams

X

[115] Proposed network architecture includes multi-interface wireless access network(eg FiWi) heterogeneous backhauling distributed cloudlets hierarchical structure ofa cloudlet and the SDN based mobile core network

IoT big datastreams

X

[110] A novel unified resource management scheme for Ethernet-based FiWi networks thatjointly allocates bandwidth for transmissions of both conventional broadband traffic andMEC data in a TDMA fashion

Mission-criticalIoT

X

[116] Introduce Mobile-IoT-Federation-as-a-Service (MIFaaS) to enable dynamic cooperationamong privatepublic local clouds of IoT devices at the edge of the cellular infrastructureThe selection of the best configuration of federated IoT cloud platforms are modeledas a coalition formation problem

Cellular IoT X

[117] Allocation of radio resources in a joint LTE and NB-IoT system based of MIFaasparadigm [116] Discovered that in handling high-end IoT data traffic a combinationbetween NB-IoT and LTE is essential in providing the needed high data rate and lowlatency

Mission-criticalIoT

X

[118] Integration of D2D communications into edge computing environment reduce transmis-sion delay and traffic load across the network

Mission-criticalIoT

X

[119] Use the theories of stochastic geometry queueing and parallel computing for provi-sioning and planning MEC networks

Communicationlatency

X

C Computation Offloading and Resource Allocation

1) Requirements Computation offloading is the mostprominent and widely discussed feature of MEC that empow-ers resource-constrained IoT devices with augmented com-putational capabilities [29] [33] This will not only prolongthe battery life of the IoT sensor nodes but also reduceE2E latency needed to run sophisticated applications In thefirst place UE has to decide whether to execute the rela-tively simple tasks locally or offload to the MEC servers(ie task model for binary offloading) [32] Secondly thedecision of computation offloading to the MEC servers can beperformed fully or partially In the partial offloading a subsetof computations is executed locally while the rest is offloadedto the MEC server by considering several factors such asusers or application preferences (eg application buffer state)radio and backhaul connections quality (ie between UEand MEC servers) UE capabilities or cloud capabilities andavailability [29]

The sole objective of the offloading policies need to be theminimization of execution delay Other critical concerns areto define the dependency of offloadable components of theapplications based on their ability to partition data (eg real-time user input has to be processed at UE without offloading)and to predict the execution time of multiple tasks Theexecution order or routines have to be carefully formulatedsince certain outcomes can be the inputs of other tasks Aspointed out in [32] the task models for partial offloading canbe represented by task-call graphs with sequential paralleland general dependencies

Although in MEC computation offloading enables power-ful cloud services at the edge level the insufficient batteryenergy at the tiny IoT devices may incur new challenges In

applications like IoT surveillance or remote asset managementthe nodes are typically hard to reach Those applications mayalso require to offload data more frequently in small chunksby consuming more energy Therefore it is necessary to con-sider not only the trade-off between energy consumption andexecution delay in both full and partial offloading scenarios inMEC but also the trade-off between computation energy andtransmission energy consumption in order to extend batterylife

The joint computation and communication resource alloca-tion should be properly addressed in order to get the maximumutilization of available resources Single MEC server will beallocated for the applications which cannot be partitionedThe resources in multiple MEC servers are allocated for theoffloaded applications that can be split into several parts Whena job arrives at the MEC server if there are enough resourcesthe scheduler has to allocate the VM for further processingIf there are no sufficient computation resources it delegatesthe task to the centralized cloud MEC servers also have toallocate computation and communication resources for userapplication jobs and MEC service jobs User mobility networktopology network scalability and load balancing are someother factors to be considered in order to define fare resourceutilization policies on MEC servers Specifically when IoTgateways share limited bandwidth among multiple IoT deviceswhich can handle video audio or bio-medical signals theallocation of bandwidth will become challenging [120] Thelow power wireless technologies (eg BLE ZigBee lowpower Wi-Fi and LPWAN standards like LoRA or SigFox)used in IoT networks have limited bandwidth When the IoTdevices access the MEC server which is acting as the IoTgateway they have to utilize either of those low-power wireless

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

9

an efficient system is required to manage the generated dataie transferring storing and analyzing such huge amounts ofdata which are collected from these smart devices Thereforecloud computing is a viable solution to these IoT-based smartgrids [90]

Generally smart grids are spanning over large geograph-ical areas They often confront bandwidth bottlenecks andcommunication delays due to poor network connectivity andvast number of devices generating data Thus the traditionalcentralized cloud architecture is not suitable for the domain ofthe smart grid since it relies heavily on centralized processing[91] Many delay sensitive smart grid applications such asfault detection isolation and service restoration or VoltVARoptimization cannot tolerate round trip delay to access cen-tralized cloud systems MEC is identified as the viable cloud

computing option to address these limitations MEC allowsthe computation to be performed closer to the data sourceMoreover the potential attack points for the grid is increasingwith the growth of ubiquitous sensor deployment Everysmart IoT device can be vulnerable to potential attacks MECprovides the opportunity to enforce security mechanism closerto the end devices As such even if an attacker gains accessto an endpoint device the attack gets no further informationbeyond the local network segment since MEC has capabilitiesto notice the intrusion and cease the accessibility [85]

I Industrial Internet

The Industrial Internet of Things (IIoT) also known asIndustry 40 [92] is an application of IoT in the domain ofmanufacturing IIoT incorporates numerous advanced commu-

TABLE VI The reviewed state-of-the-art MEC integration in different IoT applications

Ref Description Smar

tH

ome

Smar

tC

ity

Hea

lthca

re

IoT

Aut

omot

ive

Gam

ing

AR

VR

Ret

ail

Wea

rabl

eIo

T

Smar

tA

gric

ultu

re

Smar

tE

nerg

y

Indu

stri

alIn

tern

et

[40] Preliminary design of deploying MEC server functionalities in a smart hometo realize IoT gateway with direct M2M interaction in LTE networks

X

[41] Introduce Gateway-as-a-Service for heterogeneous IoT devices on top of thevirtualization technologies in edge computing

X X

[44] Propose an autonomic creation of MEC services to enhance QoS of videostreaming in smart cities

X

[50] Propose a semantic edge-based IoT architecture for military health services inbattlefield

X

[51] Provide a conceptual MEC based architecture for mission-critical context awarecollaboration in remote surgeries

X

[53] Describe and implement a smart e-Health gateway at the edge of the networksuitable for ubiquitous healthcare systems

X

[64] Analysis on research and engineering challenges co-existence of cloud edgecomputing and data caching strategies at the edge for vehicular networks

X

[82] Discuss the design aspects for the radio access in 5G V2X X[65] Discuss the benefits of merging MEC and mmWave technologies for 5G

applicationsX X X

[66] Propose a novel MEC-based architecture for future cellular vehicular networks X[67] Discuss the benefits of combining ICN and MEC in the context of connected

vehicle environmentsX

[52] Propose a framework for big data analytics between edge and cloud computingplatforms

X X

[74] Design and implement a fog computing based framework that support sharingand reusing contextual data across services in smart city and retail stores

X X

[83] Present a usecase of MEC for Tactile Internet based 5G gaming application X[84] A demonstration of MEC for Tactile Internet based 5G gaming application X[76] Discuss the role of MEC in 5G WIoT communication and its challenges X X[68] Propose an UAV-based IoT platform for a crowd surveillance use case X X[78] Develop and test a ubiquitous sensor network platform for crop lands automa-

tion maintenance in precision agricultureX

[70] Present a serverless edge computing architecture that enables the offloading ofmobile computation with low latency and high throughput using a mobile ARapplication

X

[85] Discuss the benefits of MEC and edge computing (EC) to enhance the securityof smart grids

X

[86] Present a method to optimize the EC based video streaming schemes forIndustrial IoT

X

[87] Present the use of edge computing to provide elastic resources and services toenable microdatabases architecture for IIoT

X

[88] Propose a fog-based communication architecture for Industry 40 applications X[73] Describe research directions and enablers of wireless interconnected VR

systemsX

[89] Design an optimization framework for VRAR communication via small-cellcooperation

X

10

nication and automation technologies such as M2M commu-nication machine learning and big data analytics to improveintelligence and the connectivity [93] For instance IIoTnetworks can connect all of the employees data and processesfrom the factory floor and forward them to the executiveoffices Thus decision makers or employees can create a fulland accurate view of their manufacturing process by usingIIoT network hence improving their ability to make moreinformed decisions IIoT also helps the exploitation as well asimplementation of new intelligent technologies to acceleratethe innovation and transformation of the factory workforce[92]

Primarily IIoT is seen as a way to improve operationalefficiency However IIoT provides a wide range of otherbenefits such as improving connectivity efficiency scalabilitytime savings as well as cost savings for manufacturing pro-cesses with the maximum use of smart machines [92] [94] Ingeneral these smart machines operate with higher accuracygreater efficiency and constant working capabilities than hu-mans [95] Thus IIoT has great potential for improving qualitycontrol sustainability and overall supply chain efficiency

MEC will play a vital role in enabling future IIoT applica-tions [96] by addressing the shortcomings of M2M communi-cation (eg latency resilience cost peer-to-peer connectivitysecurity) in IIoT domain [97] [98] Current market trendsalready show that edge computing will represent many im-plementation scenarios for IIoT For instance real-time edgeanalytics and enhanced edge security are two key drivers inthe creation of new IIoT deployments Thus the addition ofMEC in IIoT networks will fuel the evolution of IIoT as wellas create new business applications [99]

One way to optimize the use of conventional edge com-puting in video streaming schemes for IIoT is presented in[86] By using machine learning algorithms edge computingcan process the sensor data before transmitting to the cloudThis mitigates against the degradation of service quality of thevideo streaming Aggregation of all the sensor data to a singledata center increases latency and raises performance concernsin IIoT domain In order to solve this issue a microdatabasearchitecture is proposed for the Industrial Internet [87] It holdsthe data close to the industrial processes but also makes itavailable near the applications that can benefit from the dataEdge computing also provides elastic resources and services toenable micro-database architecture [87] A fog-based commu-nication architecture for Industry 40 applications is proposedin [88] This approach will substantially minimize the energyconsumption of the IoT nodes Edge computational capabilitiesare further used to predict future data measurements andreduce the throughput from IoT devices to the control unit

III TECHNICAL ASPECTS OF MEC ENABLED IOT

To realize the MEC exploitation for IoT applications thekey value propositions are mostly seen from the technicalparameters such as scalability communication computationoffloading and resource allocation mobility management se-curity privacy and trust management This section describesthe state-of-the-art of each of these technical parameters hence

giving a clear background against which the benefits of MECcan be envisioned

A Scalability

1) Requirements When it comes to actual deployment ofMEC platform for IoT systems scalability is a key factorto consider The compatibility of MEC servers to multiplenetwork environments is one of the factors that will driveits large scale adoption in future networks [100] The IoTenvironment will consist of hundreds of billions of sensors ac-tuators Radio-Frequency Identification (RFID)-tagged objectssoftware vehicles and embedded systems all interconnectedin a huge network of cyber-physical systems At a utility scaleconsideration these devices will be working in close collab-oration to deliver the expected services in technologies likethe smart grids virtual power plants smart homes intelligenttransportation and smart cities That being said the role ofscalability to the realization of such a hyper-connected IoTenvironment becomes more obvious The IoT environment willrequire a dynamic range of capabilities in the network space ifsuch large numbers of devices are to be supported effectively

2) Related work Currently MEC servers have been con-firmed to be compatible with LTE macro base station (eN-odeB) sites 3G Radio Network Controller (RNC) site multi-Radio Access Technology (RAT) cell aggregation site andat the edge of the core network [2] Such multi-RAT cellaggregation schemes can be implemented indoor or outdoorsettings depending on the requirements This invariably en-ables MEC to be applied to many different possible scenariosThe larger the deployment scenarios for MEC the more therange of capabilities it can handle this also translates to higherscalability for MEC-enable technologies like IoT

Designing an edge cloud network implies that an optimallocation for citing the cloud facility is first determined In[105] authors present a design optimization scheme for theMEC architecture based on link-path formulation supportedby heuristics in order to optimize the computation time forthe scheme In this approach consideration is given to bothusers and VMs mobility Hence an optimal point to installthe MEC server is determined through a tread-off betweeninstallation cost and the quality of service to be deliveredTable VII compares the reviewed state-of-the-art scalabilityfeature in MEC enabled IoT

B Communication

1) Requirements There are three main categories for thecommunication concerns about MEC [106] Wireless accesswhile offloading to the mobile edge host Backhaul accesswhile offloading to a remote cloud server Communicationamong IoT devices mobile edge host and remote cloudservers when they collaboratively execute multiple jobs Thefirst and the second categories are the most renowned on behalfof the MEC servers which are the small scale data centersdeployed by the network operators and can be co-located withthe Wireless Access Points (WAPs) In the IoT supportiveMEC systems the consumer devices may communicate with

11

TABLE VII Comparison of the reviewed state-of-the-art scalability feature in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Addressing Search

[101] Discusses the challenges in searching imposed by the burgeoning field of IoT General IoT X[94] Examines a variety of popular and innovative IoT solutions in terms of context-aware

technology perspectives to serve as a conceptual framework for context-aware productdevelopment and research in the IoT paradigm

General IoT solu-tions

X

[102] Proposes an innovative distributed architecture combining machine-to-machine industry-mature protocols (ie MQTT and CoAP) in an original way to enhance the scalabilityof gateways for the efficient IoT-cloud integration

IoT cloud inte-gration

X X

[103] Studies an implementation of edge computing which exploits transparent computing tobuild scalable IoT platforms using transparent computing

Wearable IoT X

[104] Introduces a lightweight edge gateway for the IoT architecture using container-basedvirtualization techniques

General IoT X

the MEC servers either directly or with the support of neigh-boring devices using Device-to-Device (D2D) communicationFor the third category WAPs enable access to the remote datacenters in the central cloud through backhaul links

In order to reap the maximum advantage of computationoffloading leveraged at the edge servers MEC systems needefficient communication channels Unlike the wired connec-tions in the conventional grid computing and cloud computingthe wireless access links between the mobile devices andcloud computing resources in the edge computing paradigmcan be unstable Sudden service outages may occur with theinterruption of access links The inherent challenges withwireless communication channels like multi-path fading in-terference and spectrum shortage should always be taken intoaccount for the design of MEC systems to seamlessly integratecomputation offloading and radio resource management [32]Moreover both wireless and backhaul access links have lim-ited capacities which should be properly shared among mobiledevices in a similar way as sharing the computing resources ofthe MEC server Hence having a cooperative scheme for thejoint allocation of communication and computation resourcesis important for the successful deployment of MEC [106]Redesigning both communication and networking protocolsto integrate communication infrastructures in MEC and IoTsystems is a challenging task The key focus should be onimproving the computation efficiency with respect to datatransmission

Another major requirement is to maintain interoperabilitywhile addressing heterogeneous communication technologiesthat have to be utilized in IoT and MEC paradigms in 5GThere are plenty of radio technologies that facilitate IoT Low-Power Wide Area Networks (LPWANs) (eg WCDMA LTEnarrowband IoT (NB-IoT) Wi-Fi Bluetooth Zigbee SIGFOXand LoRA) The choice of these LPWAN technologies maycreate trade-offs among signal strength operational rangethroughput and power consumption With the arrival of 5Gthe convergence of these communication technologies needsto be achieved since one network will not be fitting based onthose trade-offs

2) Related work Recently Fog-Radio Access Network (F-RAN) was introduced by Peng et al to consolidate the hetero-geneous networks into a single network architecture with 5Geven though they do not operate in the same bands to gain highspectral and operating and energy efficiency [107] Well known

Cloud Radio Access Network (C-RAN) architecture can per-form cooperative transmission across multiple edge nodes withcentralized cloud computing servers via fronthaul links [108]Although C-RAN provides high spectral efficiencies due tothe enhanced interference management capabilities with thecentralized baseband processing at the cloud it has potentiallylarge latencies F-RAN is proposed for 5G MEC deploymentsas an advanced socially aware mobile networking architectureto provide high spectral efficiency while maintaining highenergy efficiency and low latency [107] [108] Precodingdesign resource block allocation user scheduling and cellassociation are jointly designed for radio resource allocation inF-RANs in order to optimize spectral and energy efficienciesand latency performances [109] In [110] Rimal et al pro-pose a unified Time-Division Multiple Access (TDMA) basedresource management scheme for offloading traffic over Fiber-enabled Wireless (FiWi) access networks

In the envisioned 5G systems and MEC architecture bothbackhaul and wireless access links can be facilitated bymillimeter-Wave (mmW) spectrum [111] The use of mmWspectrum will enable high data rate access to MEC function-alities with low latency On the other hand MEC provideslocal computation power usefully for optimizing the perfor-mance of mmW communications In [112] [113] the authorsaddress the joint optimization of communicationcomputationresources with mmW communication They have taken the ad-vantage of blocking probabilities by considering intermittencyof mmW multi-link communications

An open source LPWAN infrastructure called OpenChirpis discussed in [114] OpenChirp which is developed usingLoRWAN allows multiple users to provision and to managebattery-powered transducers across large areas like campusesindustrial zones or cities As pointed out in [30] [115]SDN plays a vital role in improving MEC type technologiesby removing the technical shortcomings in edge computingimplementations The authors summarize the work performedfor implementing MEC based on NFV and SDN where theSDN controller manages the communication between MECservers which form a data center at the edge Table VIIIsummarizes the reviewed state-of-the-art communication is-sues and solutions in MEC enabled IoT

12

TABLE VIII Comparison of the reviewed state-of-the-art communication issues and solutions in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Commnetworkarchitecture

Commresourceallocation

[109] Performance analysis of radio resource allocation in F-RANs for edge cache andadaptive model selection to improve spectral efficiency and energy efficiency

Low latency andhigh reliability

X

[112][113]

Use of mmWave spectrum for high data rate access to MEC servers and backhaul links Low latency andhigh reliability

X

[114] An open source LPWAN infrastructure which allows multiple users to provision andmanage battery-powered transducers across large areas

LPWANinfrastructure

X

[42] A virtualized edge computing architecture with a proxy VM migration scheme tominimize traffic in the core network

IoT big datastreams

X

[115] Proposed network architecture includes multi-interface wireless access network(eg FiWi) heterogeneous backhauling distributed cloudlets hierarchical structure ofa cloudlet and the SDN based mobile core network

IoT big datastreams

X

[110] A novel unified resource management scheme for Ethernet-based FiWi networks thatjointly allocates bandwidth for transmissions of both conventional broadband traffic andMEC data in a TDMA fashion

Mission-criticalIoT

X

[116] Introduce Mobile-IoT-Federation-as-a-Service (MIFaaS) to enable dynamic cooperationamong privatepublic local clouds of IoT devices at the edge of the cellular infrastructureThe selection of the best configuration of federated IoT cloud platforms are modeledas a coalition formation problem

Cellular IoT X

[117] Allocation of radio resources in a joint LTE and NB-IoT system based of MIFaasparadigm [116] Discovered that in handling high-end IoT data traffic a combinationbetween NB-IoT and LTE is essential in providing the needed high data rate and lowlatency

Mission-criticalIoT

X

[118] Integration of D2D communications into edge computing environment reduce transmis-sion delay and traffic load across the network

Mission-criticalIoT

X

[119] Use the theories of stochastic geometry queueing and parallel computing for provi-sioning and planning MEC networks

Communicationlatency

X

C Computation Offloading and Resource Allocation

1) Requirements Computation offloading is the mostprominent and widely discussed feature of MEC that empow-ers resource-constrained IoT devices with augmented com-putational capabilities [29] [33] This will not only prolongthe battery life of the IoT sensor nodes but also reduceE2E latency needed to run sophisticated applications In thefirst place UE has to decide whether to execute the rela-tively simple tasks locally or offload to the MEC servers(ie task model for binary offloading) [32] Secondly thedecision of computation offloading to the MEC servers can beperformed fully or partially In the partial offloading a subsetof computations is executed locally while the rest is offloadedto the MEC server by considering several factors such asusers or application preferences (eg application buffer state)radio and backhaul connections quality (ie between UEand MEC servers) UE capabilities or cloud capabilities andavailability [29]

The sole objective of the offloading policies need to be theminimization of execution delay Other critical concerns areto define the dependency of offloadable components of theapplications based on their ability to partition data (eg real-time user input has to be processed at UE without offloading)and to predict the execution time of multiple tasks Theexecution order or routines have to be carefully formulatedsince certain outcomes can be the inputs of other tasks Aspointed out in [32] the task models for partial offloading canbe represented by task-call graphs with sequential paralleland general dependencies

Although in MEC computation offloading enables power-ful cloud services at the edge level the insufficient batteryenergy at the tiny IoT devices may incur new challenges In

applications like IoT surveillance or remote asset managementthe nodes are typically hard to reach Those applications mayalso require to offload data more frequently in small chunksby consuming more energy Therefore it is necessary to con-sider not only the trade-off between energy consumption andexecution delay in both full and partial offloading scenarios inMEC but also the trade-off between computation energy andtransmission energy consumption in order to extend batterylife

The joint computation and communication resource alloca-tion should be properly addressed in order to get the maximumutilization of available resources Single MEC server will beallocated for the applications which cannot be partitionedThe resources in multiple MEC servers are allocated for theoffloaded applications that can be split into several parts Whena job arrives at the MEC server if there are enough resourcesthe scheduler has to allocate the VM for further processingIf there are no sufficient computation resources it delegatesthe task to the centralized cloud MEC servers also have toallocate computation and communication resources for userapplication jobs and MEC service jobs User mobility networktopology network scalability and load balancing are someother factors to be considered in order to define fare resourceutilization policies on MEC servers Specifically when IoTgateways share limited bandwidth among multiple IoT deviceswhich can handle video audio or bio-medical signals theallocation of bandwidth will become challenging [120] Thelow power wireless technologies (eg BLE ZigBee lowpower Wi-Fi and LPWAN standards like LoRA or SigFox)used in IoT networks have limited bandwidth When the IoTdevices access the MEC server which is acting as the IoTgateway they have to utilize either of those low-power wireless

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

10

nication and automation technologies such as M2M commu-nication machine learning and big data analytics to improveintelligence and the connectivity [93] For instance IIoTnetworks can connect all of the employees data and processesfrom the factory floor and forward them to the executiveoffices Thus decision makers or employees can create a fulland accurate view of their manufacturing process by usingIIoT network hence improving their ability to make moreinformed decisions IIoT also helps the exploitation as well asimplementation of new intelligent technologies to acceleratethe innovation and transformation of the factory workforce[92]

Primarily IIoT is seen as a way to improve operationalefficiency However IIoT provides a wide range of otherbenefits such as improving connectivity efficiency scalabilitytime savings as well as cost savings for manufacturing pro-cesses with the maximum use of smart machines [92] [94] Ingeneral these smart machines operate with higher accuracygreater efficiency and constant working capabilities than hu-mans [95] Thus IIoT has great potential for improving qualitycontrol sustainability and overall supply chain efficiency

MEC will play a vital role in enabling future IIoT applica-tions [96] by addressing the shortcomings of M2M communi-cation (eg latency resilience cost peer-to-peer connectivitysecurity) in IIoT domain [97] [98] Current market trendsalready show that edge computing will represent many im-plementation scenarios for IIoT For instance real-time edgeanalytics and enhanced edge security are two key drivers inthe creation of new IIoT deployments Thus the addition ofMEC in IIoT networks will fuel the evolution of IIoT as wellas create new business applications [99]

One way to optimize the use of conventional edge com-puting in video streaming schemes for IIoT is presented in[86] By using machine learning algorithms edge computingcan process the sensor data before transmitting to the cloudThis mitigates against the degradation of service quality of thevideo streaming Aggregation of all the sensor data to a singledata center increases latency and raises performance concernsin IIoT domain In order to solve this issue a microdatabasearchitecture is proposed for the Industrial Internet [87] It holdsthe data close to the industrial processes but also makes itavailable near the applications that can benefit from the dataEdge computing also provides elastic resources and services toenable micro-database architecture [87] A fog-based commu-nication architecture for Industry 40 applications is proposedin [88] This approach will substantially minimize the energyconsumption of the IoT nodes Edge computational capabilitiesare further used to predict future data measurements andreduce the throughput from IoT devices to the control unit

III TECHNICAL ASPECTS OF MEC ENABLED IOT

To realize the MEC exploitation for IoT applications thekey value propositions are mostly seen from the technicalparameters such as scalability communication computationoffloading and resource allocation mobility management se-curity privacy and trust management This section describesthe state-of-the-art of each of these technical parameters hence

giving a clear background against which the benefits of MECcan be envisioned

A Scalability

1) Requirements When it comes to actual deployment ofMEC platform for IoT systems scalability is a key factorto consider The compatibility of MEC servers to multiplenetwork environments is one of the factors that will driveits large scale adoption in future networks [100] The IoTenvironment will consist of hundreds of billions of sensors ac-tuators Radio-Frequency Identification (RFID)-tagged objectssoftware vehicles and embedded systems all interconnectedin a huge network of cyber-physical systems At a utility scaleconsideration these devices will be working in close collab-oration to deliver the expected services in technologies likethe smart grids virtual power plants smart homes intelligenttransportation and smart cities That being said the role ofscalability to the realization of such a hyper-connected IoTenvironment becomes more obvious The IoT environment willrequire a dynamic range of capabilities in the network space ifsuch large numbers of devices are to be supported effectively

2) Related work Currently MEC servers have been con-firmed to be compatible with LTE macro base station (eN-odeB) sites 3G Radio Network Controller (RNC) site multi-Radio Access Technology (RAT) cell aggregation site andat the edge of the core network [2] Such multi-RAT cellaggregation schemes can be implemented indoor or outdoorsettings depending on the requirements This invariably en-ables MEC to be applied to many different possible scenariosThe larger the deployment scenarios for MEC the more therange of capabilities it can handle this also translates to higherscalability for MEC-enable technologies like IoT

Designing an edge cloud network implies that an optimallocation for citing the cloud facility is first determined In[105] authors present a design optimization scheme for theMEC architecture based on link-path formulation supportedby heuristics in order to optimize the computation time forthe scheme In this approach consideration is given to bothusers and VMs mobility Hence an optimal point to installthe MEC server is determined through a tread-off betweeninstallation cost and the quality of service to be deliveredTable VII compares the reviewed state-of-the-art scalabilityfeature in MEC enabled IoT

B Communication

1) Requirements There are three main categories for thecommunication concerns about MEC [106] Wireless accesswhile offloading to the mobile edge host Backhaul accesswhile offloading to a remote cloud server Communicationamong IoT devices mobile edge host and remote cloudservers when they collaboratively execute multiple jobs Thefirst and the second categories are the most renowned on behalfof the MEC servers which are the small scale data centersdeployed by the network operators and can be co-located withthe Wireless Access Points (WAPs) In the IoT supportiveMEC systems the consumer devices may communicate with

11

TABLE VII Comparison of the reviewed state-of-the-art scalability feature in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Addressing Search

[101] Discusses the challenges in searching imposed by the burgeoning field of IoT General IoT X[94] Examines a variety of popular and innovative IoT solutions in terms of context-aware

technology perspectives to serve as a conceptual framework for context-aware productdevelopment and research in the IoT paradigm

General IoT solu-tions

X

[102] Proposes an innovative distributed architecture combining machine-to-machine industry-mature protocols (ie MQTT and CoAP) in an original way to enhance the scalabilityof gateways for the efficient IoT-cloud integration

IoT cloud inte-gration

X X

[103] Studies an implementation of edge computing which exploits transparent computing tobuild scalable IoT platforms using transparent computing

Wearable IoT X

[104] Introduces a lightweight edge gateway for the IoT architecture using container-basedvirtualization techniques

General IoT X

the MEC servers either directly or with the support of neigh-boring devices using Device-to-Device (D2D) communicationFor the third category WAPs enable access to the remote datacenters in the central cloud through backhaul links

In order to reap the maximum advantage of computationoffloading leveraged at the edge servers MEC systems needefficient communication channels Unlike the wired connec-tions in the conventional grid computing and cloud computingthe wireless access links between the mobile devices andcloud computing resources in the edge computing paradigmcan be unstable Sudden service outages may occur with theinterruption of access links The inherent challenges withwireless communication channels like multi-path fading in-terference and spectrum shortage should always be taken intoaccount for the design of MEC systems to seamlessly integratecomputation offloading and radio resource management [32]Moreover both wireless and backhaul access links have lim-ited capacities which should be properly shared among mobiledevices in a similar way as sharing the computing resources ofthe MEC server Hence having a cooperative scheme for thejoint allocation of communication and computation resourcesis important for the successful deployment of MEC [106]Redesigning both communication and networking protocolsto integrate communication infrastructures in MEC and IoTsystems is a challenging task The key focus should be onimproving the computation efficiency with respect to datatransmission

Another major requirement is to maintain interoperabilitywhile addressing heterogeneous communication technologiesthat have to be utilized in IoT and MEC paradigms in 5GThere are plenty of radio technologies that facilitate IoT Low-Power Wide Area Networks (LPWANs) (eg WCDMA LTEnarrowband IoT (NB-IoT) Wi-Fi Bluetooth Zigbee SIGFOXand LoRA) The choice of these LPWAN technologies maycreate trade-offs among signal strength operational rangethroughput and power consumption With the arrival of 5Gthe convergence of these communication technologies needsto be achieved since one network will not be fitting based onthose trade-offs

2) Related work Recently Fog-Radio Access Network (F-RAN) was introduced by Peng et al to consolidate the hetero-geneous networks into a single network architecture with 5Geven though they do not operate in the same bands to gain highspectral and operating and energy efficiency [107] Well known

Cloud Radio Access Network (C-RAN) architecture can per-form cooperative transmission across multiple edge nodes withcentralized cloud computing servers via fronthaul links [108]Although C-RAN provides high spectral efficiencies due tothe enhanced interference management capabilities with thecentralized baseband processing at the cloud it has potentiallylarge latencies F-RAN is proposed for 5G MEC deploymentsas an advanced socially aware mobile networking architectureto provide high spectral efficiency while maintaining highenergy efficiency and low latency [107] [108] Precodingdesign resource block allocation user scheduling and cellassociation are jointly designed for radio resource allocation inF-RANs in order to optimize spectral and energy efficienciesand latency performances [109] In [110] Rimal et al pro-pose a unified Time-Division Multiple Access (TDMA) basedresource management scheme for offloading traffic over Fiber-enabled Wireless (FiWi) access networks

In the envisioned 5G systems and MEC architecture bothbackhaul and wireless access links can be facilitated bymillimeter-Wave (mmW) spectrum [111] The use of mmWspectrum will enable high data rate access to MEC function-alities with low latency On the other hand MEC provideslocal computation power usefully for optimizing the perfor-mance of mmW communications In [112] [113] the authorsaddress the joint optimization of communicationcomputationresources with mmW communication They have taken the ad-vantage of blocking probabilities by considering intermittencyof mmW multi-link communications

An open source LPWAN infrastructure called OpenChirpis discussed in [114] OpenChirp which is developed usingLoRWAN allows multiple users to provision and to managebattery-powered transducers across large areas like campusesindustrial zones or cities As pointed out in [30] [115]SDN plays a vital role in improving MEC type technologiesby removing the technical shortcomings in edge computingimplementations The authors summarize the work performedfor implementing MEC based on NFV and SDN where theSDN controller manages the communication between MECservers which form a data center at the edge Table VIIIsummarizes the reviewed state-of-the-art communication is-sues and solutions in MEC enabled IoT

12

TABLE VIII Comparison of the reviewed state-of-the-art communication issues and solutions in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Commnetworkarchitecture

Commresourceallocation

[109] Performance analysis of radio resource allocation in F-RANs for edge cache andadaptive model selection to improve spectral efficiency and energy efficiency

Low latency andhigh reliability

X

[112][113]

Use of mmWave spectrum for high data rate access to MEC servers and backhaul links Low latency andhigh reliability

X

[114] An open source LPWAN infrastructure which allows multiple users to provision andmanage battery-powered transducers across large areas

LPWANinfrastructure

X

[42] A virtualized edge computing architecture with a proxy VM migration scheme tominimize traffic in the core network

IoT big datastreams

X

[115] Proposed network architecture includes multi-interface wireless access network(eg FiWi) heterogeneous backhauling distributed cloudlets hierarchical structure ofa cloudlet and the SDN based mobile core network

IoT big datastreams

X

[110] A novel unified resource management scheme for Ethernet-based FiWi networks thatjointly allocates bandwidth for transmissions of both conventional broadband traffic andMEC data in a TDMA fashion

Mission-criticalIoT

X

[116] Introduce Mobile-IoT-Federation-as-a-Service (MIFaaS) to enable dynamic cooperationamong privatepublic local clouds of IoT devices at the edge of the cellular infrastructureThe selection of the best configuration of federated IoT cloud platforms are modeledas a coalition formation problem

Cellular IoT X

[117] Allocation of radio resources in a joint LTE and NB-IoT system based of MIFaasparadigm [116] Discovered that in handling high-end IoT data traffic a combinationbetween NB-IoT and LTE is essential in providing the needed high data rate and lowlatency

Mission-criticalIoT

X

[118] Integration of D2D communications into edge computing environment reduce transmis-sion delay and traffic load across the network

Mission-criticalIoT

X

[119] Use the theories of stochastic geometry queueing and parallel computing for provi-sioning and planning MEC networks

Communicationlatency

X

C Computation Offloading and Resource Allocation

1) Requirements Computation offloading is the mostprominent and widely discussed feature of MEC that empow-ers resource-constrained IoT devices with augmented com-putational capabilities [29] [33] This will not only prolongthe battery life of the IoT sensor nodes but also reduceE2E latency needed to run sophisticated applications In thefirst place UE has to decide whether to execute the rela-tively simple tasks locally or offload to the MEC servers(ie task model for binary offloading) [32] Secondly thedecision of computation offloading to the MEC servers can beperformed fully or partially In the partial offloading a subsetof computations is executed locally while the rest is offloadedto the MEC server by considering several factors such asusers or application preferences (eg application buffer state)radio and backhaul connections quality (ie between UEand MEC servers) UE capabilities or cloud capabilities andavailability [29]

The sole objective of the offloading policies need to be theminimization of execution delay Other critical concerns areto define the dependency of offloadable components of theapplications based on their ability to partition data (eg real-time user input has to be processed at UE without offloading)and to predict the execution time of multiple tasks Theexecution order or routines have to be carefully formulatedsince certain outcomes can be the inputs of other tasks Aspointed out in [32] the task models for partial offloading canbe represented by task-call graphs with sequential paralleland general dependencies

Although in MEC computation offloading enables power-ful cloud services at the edge level the insufficient batteryenergy at the tiny IoT devices may incur new challenges In

applications like IoT surveillance or remote asset managementthe nodes are typically hard to reach Those applications mayalso require to offload data more frequently in small chunksby consuming more energy Therefore it is necessary to con-sider not only the trade-off between energy consumption andexecution delay in both full and partial offloading scenarios inMEC but also the trade-off between computation energy andtransmission energy consumption in order to extend batterylife

The joint computation and communication resource alloca-tion should be properly addressed in order to get the maximumutilization of available resources Single MEC server will beallocated for the applications which cannot be partitionedThe resources in multiple MEC servers are allocated for theoffloaded applications that can be split into several parts Whena job arrives at the MEC server if there are enough resourcesthe scheduler has to allocate the VM for further processingIf there are no sufficient computation resources it delegatesthe task to the centralized cloud MEC servers also have toallocate computation and communication resources for userapplication jobs and MEC service jobs User mobility networktopology network scalability and load balancing are someother factors to be considered in order to define fare resourceutilization policies on MEC servers Specifically when IoTgateways share limited bandwidth among multiple IoT deviceswhich can handle video audio or bio-medical signals theallocation of bandwidth will become challenging [120] Thelow power wireless technologies (eg BLE ZigBee lowpower Wi-Fi and LPWAN standards like LoRA or SigFox)used in IoT networks have limited bandwidth When the IoTdevices access the MEC server which is acting as the IoTgateway they have to utilize either of those low-power wireless

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

11

TABLE VII Comparison of the reviewed state-of-the-art scalability feature in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Addressing Search

[101] Discusses the challenges in searching imposed by the burgeoning field of IoT General IoT X[94] Examines a variety of popular and innovative IoT solutions in terms of context-aware

technology perspectives to serve as a conceptual framework for context-aware productdevelopment and research in the IoT paradigm

General IoT solu-tions

X

[102] Proposes an innovative distributed architecture combining machine-to-machine industry-mature protocols (ie MQTT and CoAP) in an original way to enhance the scalabilityof gateways for the efficient IoT-cloud integration

IoT cloud inte-gration

X X

[103] Studies an implementation of edge computing which exploits transparent computing tobuild scalable IoT platforms using transparent computing

Wearable IoT X

[104] Introduces a lightweight edge gateway for the IoT architecture using container-basedvirtualization techniques

General IoT X

the MEC servers either directly or with the support of neigh-boring devices using Device-to-Device (D2D) communicationFor the third category WAPs enable access to the remote datacenters in the central cloud through backhaul links

In order to reap the maximum advantage of computationoffloading leveraged at the edge servers MEC systems needefficient communication channels Unlike the wired connec-tions in the conventional grid computing and cloud computingthe wireless access links between the mobile devices andcloud computing resources in the edge computing paradigmcan be unstable Sudden service outages may occur with theinterruption of access links The inherent challenges withwireless communication channels like multi-path fading in-terference and spectrum shortage should always be taken intoaccount for the design of MEC systems to seamlessly integratecomputation offloading and radio resource management [32]Moreover both wireless and backhaul access links have lim-ited capacities which should be properly shared among mobiledevices in a similar way as sharing the computing resources ofthe MEC server Hence having a cooperative scheme for thejoint allocation of communication and computation resourcesis important for the successful deployment of MEC [106]Redesigning both communication and networking protocolsto integrate communication infrastructures in MEC and IoTsystems is a challenging task The key focus should be onimproving the computation efficiency with respect to datatransmission

Another major requirement is to maintain interoperabilitywhile addressing heterogeneous communication technologiesthat have to be utilized in IoT and MEC paradigms in 5GThere are plenty of radio technologies that facilitate IoT Low-Power Wide Area Networks (LPWANs) (eg WCDMA LTEnarrowband IoT (NB-IoT) Wi-Fi Bluetooth Zigbee SIGFOXand LoRA) The choice of these LPWAN technologies maycreate trade-offs among signal strength operational rangethroughput and power consumption With the arrival of 5Gthe convergence of these communication technologies needsto be achieved since one network will not be fitting based onthose trade-offs

2) Related work Recently Fog-Radio Access Network (F-RAN) was introduced by Peng et al to consolidate the hetero-geneous networks into a single network architecture with 5Geven though they do not operate in the same bands to gain highspectral and operating and energy efficiency [107] Well known

Cloud Radio Access Network (C-RAN) architecture can per-form cooperative transmission across multiple edge nodes withcentralized cloud computing servers via fronthaul links [108]Although C-RAN provides high spectral efficiencies due tothe enhanced interference management capabilities with thecentralized baseband processing at the cloud it has potentiallylarge latencies F-RAN is proposed for 5G MEC deploymentsas an advanced socially aware mobile networking architectureto provide high spectral efficiency while maintaining highenergy efficiency and low latency [107] [108] Precodingdesign resource block allocation user scheduling and cellassociation are jointly designed for radio resource allocation inF-RANs in order to optimize spectral and energy efficienciesand latency performances [109] In [110] Rimal et al pro-pose a unified Time-Division Multiple Access (TDMA) basedresource management scheme for offloading traffic over Fiber-enabled Wireless (FiWi) access networks

In the envisioned 5G systems and MEC architecture bothbackhaul and wireless access links can be facilitated bymillimeter-Wave (mmW) spectrum [111] The use of mmWspectrum will enable high data rate access to MEC function-alities with low latency On the other hand MEC provideslocal computation power usefully for optimizing the perfor-mance of mmW communications In [112] [113] the authorsaddress the joint optimization of communicationcomputationresources with mmW communication They have taken the ad-vantage of blocking probabilities by considering intermittencyof mmW multi-link communications

An open source LPWAN infrastructure called OpenChirpis discussed in [114] OpenChirp which is developed usingLoRWAN allows multiple users to provision and to managebattery-powered transducers across large areas like campusesindustrial zones or cities As pointed out in [30] [115]SDN plays a vital role in improving MEC type technologiesby removing the technical shortcomings in edge computingimplementations The authors summarize the work performedfor implementing MEC based on NFV and SDN where theSDN controller manages the communication between MECservers which form a data center at the edge Table VIIIsummarizes the reviewed state-of-the-art communication is-sues and solutions in MEC enabled IoT

12

TABLE VIII Comparison of the reviewed state-of-the-art communication issues and solutions in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Commnetworkarchitecture

Commresourceallocation

[109] Performance analysis of radio resource allocation in F-RANs for edge cache andadaptive model selection to improve spectral efficiency and energy efficiency

Low latency andhigh reliability

X

[112][113]

Use of mmWave spectrum for high data rate access to MEC servers and backhaul links Low latency andhigh reliability

X

[114] An open source LPWAN infrastructure which allows multiple users to provision andmanage battery-powered transducers across large areas

LPWANinfrastructure

X

[42] A virtualized edge computing architecture with a proxy VM migration scheme tominimize traffic in the core network

IoT big datastreams

X

[115] Proposed network architecture includes multi-interface wireless access network(eg FiWi) heterogeneous backhauling distributed cloudlets hierarchical structure ofa cloudlet and the SDN based mobile core network

IoT big datastreams

X

[110] A novel unified resource management scheme for Ethernet-based FiWi networks thatjointly allocates bandwidth for transmissions of both conventional broadband traffic andMEC data in a TDMA fashion

Mission-criticalIoT

X

[116] Introduce Mobile-IoT-Federation-as-a-Service (MIFaaS) to enable dynamic cooperationamong privatepublic local clouds of IoT devices at the edge of the cellular infrastructureThe selection of the best configuration of federated IoT cloud platforms are modeledas a coalition formation problem

Cellular IoT X

[117] Allocation of radio resources in a joint LTE and NB-IoT system based of MIFaasparadigm [116] Discovered that in handling high-end IoT data traffic a combinationbetween NB-IoT and LTE is essential in providing the needed high data rate and lowlatency

Mission-criticalIoT

X

[118] Integration of D2D communications into edge computing environment reduce transmis-sion delay and traffic load across the network

Mission-criticalIoT

X

[119] Use the theories of stochastic geometry queueing and parallel computing for provi-sioning and planning MEC networks

Communicationlatency

X

C Computation Offloading and Resource Allocation

1) Requirements Computation offloading is the mostprominent and widely discussed feature of MEC that empow-ers resource-constrained IoT devices with augmented com-putational capabilities [29] [33] This will not only prolongthe battery life of the IoT sensor nodes but also reduceE2E latency needed to run sophisticated applications In thefirst place UE has to decide whether to execute the rela-tively simple tasks locally or offload to the MEC servers(ie task model for binary offloading) [32] Secondly thedecision of computation offloading to the MEC servers can beperformed fully or partially In the partial offloading a subsetof computations is executed locally while the rest is offloadedto the MEC server by considering several factors such asusers or application preferences (eg application buffer state)radio and backhaul connections quality (ie between UEand MEC servers) UE capabilities or cloud capabilities andavailability [29]

The sole objective of the offloading policies need to be theminimization of execution delay Other critical concerns areto define the dependency of offloadable components of theapplications based on their ability to partition data (eg real-time user input has to be processed at UE without offloading)and to predict the execution time of multiple tasks Theexecution order or routines have to be carefully formulatedsince certain outcomes can be the inputs of other tasks Aspointed out in [32] the task models for partial offloading canbe represented by task-call graphs with sequential paralleland general dependencies

Although in MEC computation offloading enables power-ful cloud services at the edge level the insufficient batteryenergy at the tiny IoT devices may incur new challenges In

applications like IoT surveillance or remote asset managementthe nodes are typically hard to reach Those applications mayalso require to offload data more frequently in small chunksby consuming more energy Therefore it is necessary to con-sider not only the trade-off between energy consumption andexecution delay in both full and partial offloading scenarios inMEC but also the trade-off between computation energy andtransmission energy consumption in order to extend batterylife

The joint computation and communication resource alloca-tion should be properly addressed in order to get the maximumutilization of available resources Single MEC server will beallocated for the applications which cannot be partitionedThe resources in multiple MEC servers are allocated for theoffloaded applications that can be split into several parts Whena job arrives at the MEC server if there are enough resourcesthe scheduler has to allocate the VM for further processingIf there are no sufficient computation resources it delegatesthe task to the centralized cloud MEC servers also have toallocate computation and communication resources for userapplication jobs and MEC service jobs User mobility networktopology network scalability and load balancing are someother factors to be considered in order to define fare resourceutilization policies on MEC servers Specifically when IoTgateways share limited bandwidth among multiple IoT deviceswhich can handle video audio or bio-medical signals theallocation of bandwidth will become challenging [120] Thelow power wireless technologies (eg BLE ZigBee lowpower Wi-Fi and LPWAN standards like LoRA or SigFox)used in IoT networks have limited bandwidth When the IoTdevices access the MEC server which is acting as the IoTgateway they have to utilize either of those low-power wireless

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

12

TABLE VIII Comparison of the reviewed state-of-the-art communication issues and solutions in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Commnetworkarchitecture

Commresourceallocation

[109] Performance analysis of radio resource allocation in F-RANs for edge cache andadaptive model selection to improve spectral efficiency and energy efficiency

Low latency andhigh reliability

X

[112][113]

Use of mmWave spectrum for high data rate access to MEC servers and backhaul links Low latency andhigh reliability

X

[114] An open source LPWAN infrastructure which allows multiple users to provision andmanage battery-powered transducers across large areas

LPWANinfrastructure

X

[42] A virtualized edge computing architecture with a proxy VM migration scheme tominimize traffic in the core network

IoT big datastreams

X

[115] Proposed network architecture includes multi-interface wireless access network(eg FiWi) heterogeneous backhauling distributed cloudlets hierarchical structure ofa cloudlet and the SDN based mobile core network

IoT big datastreams

X

[110] A novel unified resource management scheme for Ethernet-based FiWi networks thatjointly allocates bandwidth for transmissions of both conventional broadband traffic andMEC data in a TDMA fashion

Mission-criticalIoT

X

[116] Introduce Mobile-IoT-Federation-as-a-Service (MIFaaS) to enable dynamic cooperationamong privatepublic local clouds of IoT devices at the edge of the cellular infrastructureThe selection of the best configuration of federated IoT cloud platforms are modeledas a coalition formation problem

Cellular IoT X

[117] Allocation of radio resources in a joint LTE and NB-IoT system based of MIFaasparadigm [116] Discovered that in handling high-end IoT data traffic a combinationbetween NB-IoT and LTE is essential in providing the needed high data rate and lowlatency

Mission-criticalIoT

X

[118] Integration of D2D communications into edge computing environment reduce transmis-sion delay and traffic load across the network

Mission-criticalIoT

X

[119] Use the theories of stochastic geometry queueing and parallel computing for provi-sioning and planning MEC networks

Communicationlatency

X

C Computation Offloading and Resource Allocation

1) Requirements Computation offloading is the mostprominent and widely discussed feature of MEC that empow-ers resource-constrained IoT devices with augmented com-putational capabilities [29] [33] This will not only prolongthe battery life of the IoT sensor nodes but also reduceE2E latency needed to run sophisticated applications In thefirst place UE has to decide whether to execute the rela-tively simple tasks locally or offload to the MEC servers(ie task model for binary offloading) [32] Secondly thedecision of computation offloading to the MEC servers can beperformed fully or partially In the partial offloading a subsetof computations is executed locally while the rest is offloadedto the MEC server by considering several factors such asusers or application preferences (eg application buffer state)radio and backhaul connections quality (ie between UEand MEC servers) UE capabilities or cloud capabilities andavailability [29]

The sole objective of the offloading policies need to be theminimization of execution delay Other critical concerns areto define the dependency of offloadable components of theapplications based on their ability to partition data (eg real-time user input has to be processed at UE without offloading)and to predict the execution time of multiple tasks Theexecution order or routines have to be carefully formulatedsince certain outcomes can be the inputs of other tasks Aspointed out in [32] the task models for partial offloading canbe represented by task-call graphs with sequential paralleland general dependencies

Although in MEC computation offloading enables power-ful cloud services at the edge level the insufficient batteryenergy at the tiny IoT devices may incur new challenges In

applications like IoT surveillance or remote asset managementthe nodes are typically hard to reach Those applications mayalso require to offload data more frequently in small chunksby consuming more energy Therefore it is necessary to con-sider not only the trade-off between energy consumption andexecution delay in both full and partial offloading scenarios inMEC but also the trade-off between computation energy andtransmission energy consumption in order to extend batterylife

The joint computation and communication resource alloca-tion should be properly addressed in order to get the maximumutilization of available resources Single MEC server will beallocated for the applications which cannot be partitionedThe resources in multiple MEC servers are allocated for theoffloaded applications that can be split into several parts Whena job arrives at the MEC server if there are enough resourcesthe scheduler has to allocate the VM for further processingIf there are no sufficient computation resources it delegatesthe task to the centralized cloud MEC servers also have toallocate computation and communication resources for userapplication jobs and MEC service jobs User mobility networktopology network scalability and load balancing are someother factors to be considered in order to define fare resourceutilization policies on MEC servers Specifically when IoTgateways share limited bandwidth among multiple IoT deviceswhich can handle video audio or bio-medical signals theallocation of bandwidth will become challenging [120] Thelow power wireless technologies (eg BLE ZigBee lowpower Wi-Fi and LPWAN standards like LoRA or SigFox)used in IoT networks have limited bandwidth When the IoTdevices access the MEC server which is acting as the IoTgateway they have to utilize either of those low-power wireless

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

13

connections that have low bandwidth

2) Related work In the comprehensive survey presentedin [29] the existing work that addresses MEC computationoffloading decisions have been nicely summarized based onfull and partial offloading types These solutions are proposedeither to minimize the execution delay or to balance the trade-off between energy consumption and latency Moreover [29]provides an overview of the latest research works that addressthe allocation of computation resources for the data or appli-cation which it decides to offload in MEC systems Howeverthis analysis does not address the explicit applicability of com-putation offloading and resource allocation in IoT supportiveMEC systems

A preliminary study on how computation offloading andbandwidth allocation can be performed in MEC supportive IoTnetworks is presented in [120] Due to the discrete and coarse-grained offloading levels on the IoT end nodes the gateway(ie MEC server) bandwidth will be under-utilized Thisphenomenon is termed fragmentation Based on the receivedtransmission rates and power consumption parameters of IoTdevices the gateway runs an iterative algorithm to optimallyallocate bandwidth in such a way as to optimize the batterylife of the devices The implementation of the algorithmfor a health monitoring application shows more than 40improvement in using gateway bandwidth and up to 15 hourimprovement in battery life of IoT devices Replisom [121]designed by Abdelwahab et al is a model for computationoffloading for massive IoT applications where the replicatedmemory objects produced by IoT devices are offloaded tothe LTE-aware edge cloud Replisom protocol relies on D2Dcommunication for effectively scheduling the memory repli-cation occasions to resolve interference and scarcity in radioresources as a large number of devices simultaneously transmittheir memory replicas

Furthermore with the advent of mobile device performanceand D2D communication technologies computation offloadingcan be performed at the mobile devices As shown in [129]a collection of co-located mobile devices can be utilized toprovide cloud services at the edge instead of using MECservers Such an offloading mechanism will allow the veryconstrained tiny IoT devices to outsource the computationintensive tasks to the high performing mobile devices in theclosest proximity Few research efforts were performed toderive computation offloading strategies in MEC that supportuser mobility In [130] the authors propose a hybrid compu-tation offloading mechanism for edge computing consideringthe hardware heterogeneity of the mobile devices varioususers requirements on Quality of Experience (QoE) and theheterogeneity status of the network

The requests for computation offloading generated by enddevices have to be handled by the software load balanceraccording to the availability of the MEC servers and resourcesYu et al proposes a softwarized load balancer techniquecalled SDLB for edge computing based on the minimal perfecthashing algorithm [122] Their scalable and dynamic loadbalancer SDLB is derived based on POG data structure andable to support about one million update requests per secondIn [123] the authors propose a virtualized network architecturewith intelligent resource allocation capabilities for NFV MECand IoT services This so called TelcoFog architecture providesseamless and unified control for the complete visibility com-putation and allocation of both cloud and network resourcesthrough different network segments (access aggregation andtransport) assuming heterogeneous access and transport tech-nologies (eg Wi-Fi packet switching optical transmission)

The game theoretic approach is also designed for selectingthe most appropriate wireless channels to transmit offloadingdata in a multi-user multi-channel MEC systems [131] [132]

TABLE IX Comparison of the reviewed state-of-the-art computation offloading and resource allocation features in MECenabled IoT

Ref Description IoT applicationdomainfeature

ComputationOffloading

CompResourceAllocation

[120] Management of computation offloading in a local IoT network with the efficientutilization of IoT gateway bandwidth constraints

IoT-gateway X

[121] Replicated memory objects produced by IoT devices are offloaded to the LTE-awareedge cloud based on D2D communication

Massive-IoT X

[122] Proposes a portable MEC load balancer which is scalable software based memoryefficient and adaptive to device heterogeneity The design takes the advantages of SDNand POG data structure

IoT big datastreams

X

[123] Defines an architecture to allocate cloud and edge resources for deploying NFV MECand IoT services on top of a telecom operatorrsquos network

Low latency X

[124] Propose a MEC clustering algorithm to consolidate the maximum communications atthe edge which stands for the spatial temporal dynamics of the traffic

IoT big datastreams

X

[125] Defines a scalable offloading architecture and a simulator with multi-tenancy ability anddynamic horizontal scaling based on Amazon Autoscale service-oriented architecture

Massive-IoT X X

[126] Formulate the computation offloading decision resource allocation and content cachingin wireless cellular networks with mobile edge computing as an optimization problemand solve it applying alternating direction method of multipliers based distributedalgorithm

Cellular IoT X X

[127] Introduces asymptotically optimal offloading schedules which are tolerant to partialout-of-date network knowledge and stochastically maximize a time-average networkutility balancing system throughput and fairness

Massive IoT X X

[128] Develop a toolkit for modeling and simulation of resource management techniques inthe IoT edge and fog computing environments

General IoT X

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

14

In [133] the MEC server makes the offloading decisions andphysical resource block allocation to the UEs using the graphcoloring method Furthermore in [124] authors propose agraph-based algorithm that takes into account the maximumMEC server capacity provides a partition of geographic areaand consolidates as many communications as possible at theedge The offloading architecture proposed in [125] addressesthe scaling of offloading support to large-scale IoT environ-ments Their application level task scheduler uses horizontalscaling to allocate the available resources in the edge cloudMoreover content caching strategy is also considered in somework for the optimized joint computation and communicationresource allocation [126] Table IX summarizes the reviewedstate-of-the-art computation offloading and resource allocationfeatures in MEC enabled IoT

D Mobility Management1) Requirements A more general concept in cellular and

IP networks is mobility management for moving users Sinceearlier generations of mobile cellular networks mobility man-agement has been the ultimate way of ensuring that mo-bile services are delivered to subscribers wherever they arewithin the coverage areas of the service provider The cellularnetwork is a radio network that consists of multiple basestations each base station is designated to provide mobileservices within a particular cell and hence combining severalbase stations enables the service provider to cover wider ge-ographical locations In LTE mobility management advancedsignificantly through the introduction of moving networksseamless roaming and vertical handovers which is enabledwhen the UE changes the serving eNBSCeNB

In the case of MEC mobility management is particularlycrucial given that when mobile UEs move far away from thecomputing node then there is the possibility of degradingthe QoS due to latency A severe degradation could lead toa complete disconnection of a UE from the MEC network InMEC-enabled IoT a large majority of the nodes will be mobilenodes hence the goal is to exploit MEC services to offeran ultra-reliable mobility management scheme for IoT appli-cations In traditional mobile networks the key issues withmobility management are mainly connectivity location man-agement routing group formation seamless mobility mobilitycontext management and migration among others Amongthese issues seamless mobility tends to be the most trivialThere is a need for mobile devices to have uninterrupted accessto information communication monitoring and control whenwhere and how they want regardless of the device servicenetwork or location For the MEC architecture using suchtraditional approach to mobility management will certainlylead to a degraded performance in the overall MEC networkone key reason for this shortfall is due to the co-provision ofradio access and computing services of the MEC-enabled basestations

2) Related Work Several mobility management policieshave been proposed for the MEC architecture [29] [134]ndash[136] In [134] authors developed a novel user-centric Energy-aware Mobility Management (EMM) scheme based on Lya-punov optimization and multi-armed bandit theories The

EMM scheme works in an online fashion without using futuresystem state information is hence able to manage the imperfectsystem state information The goal of EMM is to optimizethe offloading delay that results from both radio access andcomputation under the long-term energy consumption con-straint of the user Here the experiment results showed thatthe proposed algorithms can optimize the delay performancewhile approximately satisfying the energy consumption budgetof the user However a major issue with this algorithm is thatit will not be effective for a high mobility scenario where aconnected node will move in a great deal during the processingof a task and such high mobility scenario is a typical featureof the IoT networks

In [29] authors presented a user-oriented use case ofMEC from the perspective of computational offloading andmobility management They first discuss the power controlapproach where the mobility management entity regulates thetransmission power of the eNBSCeNB which is mostly usedin scenarios where the UEs mobility is confined within agiven space such as an office room [29] [137] [138] Theprinciple of this approach is depicted in Figure 4 Accordinglythe MEC services are extended to slowly moving IoT deviceswithin a given space by adjusting the transmission power of theserving andor neighboring SCeNBs This Cloud-aware PowerControl (CaPC) algorithm is mostly suitable for managing theoffloading of real-time applications where delay requirementsare strict It allows the MEC system to handle higher amountsof offloaded applications within specific latency constraintTypically increasing the transmission power of SCeNB willmomentarily increase the coverage region of MEC signalshence allowing IoT nodes to move beyond the default coverageregion for the duration of the power boost This will help toavoid the need for handover as much as possible especially incases where the moving distance of the IoT device is relativelysmall The moving IoT devices are able to roam certaindistance away from the coverage region of MEC servicesjust by adapting the transmission power of the eNBSCeNBwithout discontinuity in service and handovers

Fig 4 CaPC Power Control Principle [29]

Another scenario is when the IoT node decides to initiate anoffload either within the coverage region increased by power

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

15

TABLE X Summary of the reviewed state-of-the-art mobility management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

MobilityManage-ment

FlowSchedul-ing

[134] Develop a user-centric energy-aware mobility management (EMM) scheme to optimizethe delay due to both radio access and computation under the long-term energyconsumption constraint of the user

General IoT X

[140] Present UbiFlow the first software-defined IoT system for ubiquitous flow control andmobility management base don distributed controller in multinetworks

Software definedIoT

X X

[141] Explores how Named Data Networking a proposed future Internet architecture canaddress the challenges of interoperability in IoT networks

IoT applications X

[142] Analyzed distributed mobility management for future IoT sensor networks IoT sensors X X[143] Propose a location-aware load prediction at edge data centers which supports user

mobilityGeneral IoT X

control or as it roams beyond Two possible procedures couldbe used in this case one is by performing a VM migrationie migrating a VM from the less effective to a more effectivecomputing node and two is by path selection ie selectinga new path for communication between the computing nodeand the IoT device The need for VM migration arises whenthe IoT node roams beyond the region extended by thepower control mechanism In that case the risk of servicediscontinuity and poor QoS factors tend to be higher hencethere is a need to strategically design the VM migrationprocess Analysis of the influence of such migration on theperformance of a typical IoT node is described in [139] usingthe Markov chain analytical models Based on the outcomeof the analysis when VM migration is not implemented theprobability that the edge device will connect to the optimalMEC decreases with the increase in hops between the eNBand the UE Meanwhile there is also an additional delay thatoccurs in when VM migration is not used In addition to theliterature mentioned in [29] Table X summarizes the reviewedstate-of-the-art mobility management in MEC enabled IoT

E Security

1) Requirements Integrating MEC capabilities to the IoTsystems come with an assurance of better performance interms of quality of service and ease of implementation Thishowever raises concerns in both research and the industryfirst on the heterogeneity of connected devices and second onthe potential repercussions of such architectural modificationon the overall security of MEC-enabled systems Typicalsecurity threats in these areas are Denial of Service (DoS)attacks Man-in-the-Middle (MitM) attacks and malicious nodeproblems [144] [145] More detailed descriptions of thesethreats are presented in [145]

IoT systems in general inherit most of the security vulnera-bilities commonly found on sensor networks mobile commu-nication networks and the Internet as a whole Thus makingsecurity one of the application challenges of IoT in present andfuture networks Such security vulnerabilities in IoT networksinclude DoSDistributed DoS (DDoS) attacks forgerymiddleattack heterogeneous network attacks application risk ofIPv6 Wireless Local Area Networking (WLAN) applicationconflicts also affect the transport security of IoT [146]

Here we define the possible security attacks in the contextof MEC-enabled IoT environment Security threats are mostly

targeted towards the MEC nodes eg MEC server and otherIoT nodes In DoS attacks the adversaries tend to attack criti-cal networking or computing resources by sending requests atrates that are beyond the handling capacity of such networkingor computing equipment hence inundating such facility andpreventing other users or nodes from getting access to theresources offered DoS attacks could happen in the form ofDDoS or wireless jamming and could be launched on boththe virtualization and network infrastructures

MitM happens when an adversary interposes between twonodes or entities and secretly relaying or altering the com-munication between such parties common example is theMitM attack between a server and a client For the MEC-enabled IoT scenario the most vulnerable location for MitMattack is the infrastructure layer where the malicious attackertries to hijack certain segments of the network and begins tolaunch attacks like eavesdropping and phishing on connecteddevices As claimed in [147] MitM attacks can be launchedbetween 3G and WLAN networks Such attacks would be evenmore threatening for the MEC-enabled IoT scenario giventhat MEC relies heavily on virtualization hence launching aMitM attack on multiple VMs could very easily affect all otherelements on both sides of the attack

VM Manipulation is a typical attack for all virtualizedand edge computing systems In MEC-enabled IoT systemVM manipulation is mainly targeted towards the virtualizationinfrastructures In this case the attacker is more likely to bea malicious insider with enough privileges or a VM that hasescalated privileges The adversary in such attack begins tolaunch multiple attacks to the VMs running inside it WhenVM manipulation attack is launched the affected VMs arefurther exposed to numerous other potential attacks like logicbombs

2) Related Work On the application layer security threatsare mostly in the context of information access and userauthentication Others include possibility of tracking and de-stroying data streams tampering with the stability of the IoTplatform attacking the middleware layer andor managementplatform [148] [149] Given that IoT will further convergepeoples everyday life activities and devices on the networkthe need for faster access to data which is largely addressedby introducing MEC to the IoT system must be balanced bya robust and highly reliable security technology in additionto creating more security awareness for users and applicationdevelopers

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

16

The architecture proposed in [146] has three key layersnamely perception transportation and application The authorshave identified different potential security vulnerabilities oneach layer For the perception layer potential security vulner-abilities are mainly on the RFID the wireless sensor networksand the RFID sensor networks For the transport layer securityvulnerabilities are mainly found at the access network thecore network and the local network Here vulnerabilitiescan also be unique to the different access technologies iefor 3G access network Ad-Hoc network and Wi-Fi Onthe application layer vulnerabilities exist for the applicationsupport layer as well as for specific IoT applications

F Privacy1) Issues and challenges The early designs of IoT systems

were largely closed homogeneous and single-purpose withlimited functionality geographic scope and scale In contrastthe present-day IoT systems are much larger and spanningacross countries or continents making them to comply withthe varying rules and regulations Similarly in health care[150] type of applications which invade personal spaces pri-vacy is becoming a significant concern [10] [12] Governingorganizations like European Commission have recognized thatprivacy in the processing of personal data and the confiden-tiality of communications as fundamental rights that should beprotected [151] In an IoT application when the data sharingprinciple is leveraged by a cloud based system that couldraise a lot of privacy concerns The potential use of data forunpredicted future applications may compromise privacy

MEC enables caching data processing and analytics to bedone closer the source of the data and reduces the burden oncentralized cloud servers and core networks [22] Importantlythis will support differentiated privacy since raw unprocesseddata does not have to be stored or processed by a centralizedcloud systems which can be located in distance Only theprocessed and selected data are needed to reach the centralizecloud for further processing [10] [12] For instance the imageprocessing of car number plate recognition can be done inthe edge without transferring the location information to thecentralized cloud servers Such MEC based local processingprotects the privacy of data without leaving the jurisdictionof the user Moreover the decentralized approach reducesthe impact of data breaches such as Sony breach [152] andOPM (Office of Personnel Management) breach [153] MECapproach also enable the possibility to implement specific orlocal privacy policies [154] contrary to the uniform privacypolicies applied in centrally managed public cloud In someIoT applications such eHealth services (for instance mentaland abortion clinics) local privacy polices with edge intelli-gence is required to meet the required privacy protection whichcannot be met by only using a centralized approach [154]

The requirements in privacy protection are identified basedon the generic and the regulatory objectives First it is requiredto harmonize the privacy of digital services at global level bypromoting the digital single market All relevant directives andlegislative instruments should be encouraged to enable crossborder policies Then it is necessary to balance the interests inprotecting privacy and in fostering the global use of services

Second the privacy legislation should be done at a globallevel to ensure their compatibility with new technologies suchas MEC Different jurisdictions should cooperate togetherto develop inter-operable privacy requirements and facilitatethe flow of information with the required level of privacyprotection For instance the rdquoSafe Harborrdquo agreement betweenUS and EU requires US companies to obey EU regulationsso that EU companies can store and process data in US datacenters [155]

Third it is necessary to foster interoperability and dataportability to support the adaptation of new technologies Forinstance it can be done by avoiding mandated standards orpreferences which could prevent interoperability Moreover itis necessary to promote the on-going interoperability effortsin the industries this will be useful in defining uniform andglobal privacy policies Finally it is required to define oneframework with a set of data protection laws which can beused across the border and they should be simple enough tobe set up globally This framework should be based on theconcept of accountability and the laws should also supportself-regulatory codes and mechanisms

2) Related work Security and privacy challenges in MEClike edge computing paradigms are surveyed in [34] and [156]A partially distributed approach that allows edge intelligencethat can meet the privacy requirements of IoT use cases suchas eHealth services is presented in [154] The possibility ofexploiting edge computing to solve the problem of loss ofprivacy by releasing personal and social data to centralizedservices such as e-commerce sites rating services searchengines social networks and location services are presentedin [22] Possibilities of improving the data privacy of IoT databy using edge computing is presented in [45]

G Trust management

1) Requirements Trust is a rather complex property todefine it is closely associated with the overall security of anynetwork or platform Trust is significant in critical 5G usecases like remote surgeries emergency autonomous vehiclesfactory automation and tele-operated driving (eg drones) Inthese scenarios latency and reliability are highly regardedAlthough trust is an equally important property similar tosecurity and privacy in IoT and MEC it is hardly addressedlately in research works [34] The need to implement theappropriate trust management scheme is very essential when itcomes to IoT technologies This is because IoT devices offloadtheir delay critical applications to the edge cloud which isnormally out of the direct control of the client

According to Yan et al the key challenges of trustmanagement in IoT are not only limited to system securityrobustness and privacy preservation [13] Trust relationshipshave to be sustained among all IoT system entities includingthe enabling technologies such as MEC Data perception trustdetermines the reliability of data sensing and collection in theIoT perception layer Data fusion and mining trust explainsthe efficiency and trustworthiness of big data handling inthe IoT network layer Enabling secure data transmission andcommunication while maintaining the quality of IoT services

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

17

and identity trust are other important aspects of IoT trust It isequally important to apply a more generic trust managementframework for IoT since it is a collaboration of multipletechnologies and systems The utilization of tamper resistivesecure elements will enable the trust in the end user deviceswith physical protections to prevent the compromising ofcryptographic security parameters However due to limitedresources in many tiny IoT devices the integration of suchtrust enabling devices will also be challenging Above all themost significant is the realization of human-computer trustinteraction which requires more attention to the subjectiveproperties of IoT users at the application layer

In cloud computing trust is targeted towards long-termunderlying properties or infrastructure (persistent trust) andsuch trust can be specific to context-based social and tech-nological mechanisms (dynamic trust) Moreover when edgecloud computing is collaborating with IoT it introduces moretrust related objectives such as maintaining the trust forcomputation offloading IoT services or collected data to theedge cloud and the cooperative trust among edge servers Theedge servers should ensure the trustworthiness of end users andIoT devices which acquire the resources from the edge cloudLikewise the edge servers should also assure their reliabilityand trustworthiness to the end usersdevices and other edgeservers for providing guaranteed services More importantlythe efficient resource sharing among the edge servers has to beaccomplished based on a proper trust management framework

2) Related work The comprehensive literature surveysin [10] [13] summarize the recent research works on IoTtrust Accordingly the researchers have addressed IoT trust inmultiple perspectives including trust evaluation trust frame-work data perception trust identity trust and privacy preser-vation transmission and communication trust secure multi-party computation user trust and application trust ExistingIoT trust evaluation mechanisms are mathematically formed

and have considered different trust metrics like social trustand QoS trust using both direct observations and indirectrecommendations Most of the trust frameworks proposed inIoT address security and privacy in IoT data transmission andcommunications In [159] a preliminary design of a holisticsolution with trust and security-by-design for cyber physicalsystems based on IoT and cloud architectures is presentedThey have taken the initiative to develop and demonstrate atrustworthy-by-design autonomic security framework based onSDNNFV and IoT networks

In many previous literatures data perception trust is ad-dressed in the context of security and privacy mainly bymitigating security attacks on data aggregation and processingas well as exploiting some key management techniques [13]Some recent literatures have also addressed data protectionand performance improvement at the edge computing serversby trust management among fog servers [160] Furthermoretrust is paramount to the effectiveness of node interaction inSIoT where the objects are building up a social network andbecoming more autonomous [14] Table XI summarizes the re-viewed state-of-the-art security privacy and trust managementin MEC enabled IoT

IV INTEGRATION TECHNOLOGIES

The realization of MEC for IoT is fueled by several in-tegrating technologies such as SDN NFV ICN and NetworkSlicing This section provides a high level overview of the roleof each technology in MEC-IoT environment and the relatedworks

A Network Function Virtualization

NFV is a network concept which proposes to use virtualiza-tion technologies to manage core networking functions usinga software based approach [161] NFV has been proven as one

TABLE XI Comparison of the reviewed state-of-the-art security privacy and trust management in MEC enabled IoT

Ref Description IoT applicationdomainfeature

Security Privacy Trust

[144] Proposed a security framework for virtualized Small Cell Networks with theaim of further extending MEC in the broader 5G environment

Cloud-enabled IoT X X

[157] Addresses the utility based matching or pairing problem within the samedomain of IoT nodes by using Irvingrsquos matching algorithm under the nodespecified preferences to endure a stable IoT node pairing

IoT node pairingservices

X X X

[146] Analyzes the cross-layer heterogeneous integration issues and security issuesin detail and discusses the security issues of IoT as a whole and tries to findsolutions to them

General IoT X X

[22] Presents the research challenges associated with security privacy and trustmanagement in Edge-centric Computing

General IoT X X X

[158] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[34] Holistically analyses the security and privacy threats challenges and mecha-nisms inherent in all edge paradigms including MEC

General IoT X X

[156] A survey on security and privacy challenge in fog computing General IoT X X[154] Present a edge computing based distributed approach to satisfy the security and

privacy requirements of IoTGeneral IoT X X

[45] Discuss the methods of improving security and privacy of IoT data by usingedge computing

General IoT X X

[159] Introduce the preliminary design of a holistic framework for enabling trust andsecurity by-design for cyber physical systems (CPS) based on IoT and edgecloud architectures

IoT architecture X X X

[160] Propose a trust translation model for fog nodes and a privacy-aware model foraccess control at fog nodes

IoT big datastreams

X X

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

18

of the key enablers for not only the development of 5G butalso MEC-IoT integration [162] Specifically MEC reuses theNFV virtualization infrastructure and the NFV infrastructuremanagement to the largest extent possible [163]

Both MEC and NFV technologies can be used together inenvironments such as 5G mobile networks to elevate com-puting capacity to meet the increased networking demandsMEC architecture is also based on a virtualized platformquite similar to NFV architecture Both technologies featurestackable components and each has a virtualization layer

According to ESTI [2] it is beneficial to reuse the in-frastructure and infrastructure management of NFV to thelargest extent possible by hosting both Virtual Network Func-tions (VNFs) and MEC applications on the same platformcomputing experience is enhanced The use of NFV willequally increase the scalability of MEC application NFV canimprove the scalability by dynamically scaling updown thenetwork resources depending on demand

Several NFV-MEC ingratiation research works have beenproposed recently In [163] NFV-enabled MEC scheme isproposed to optimize the placement of resources among NFV-enabled nodes to support low latency mobile multimediaapplications A novel MEC and NFV integrated networkarchitecture is presented in [164] this can be used to enhancethe mobile game experience optimized high speed HD videostreaming and local content caching for AR The double-tier MEC-NFV architecture in [165] aligns and integratesthe MEC system with the NFV Management and Orchestra-tion (MANO) by introducing a management subsystem thatenriches the MANO with application-oriented orchestrationcapabilities To support the deployment of container-basednetwork services at the edge of the network an architecturebased on the Open Baton MANO framework is proposed bycombining the NFV and MEC within a single orchestrationenvironment [166]

B Software Defined Networking

SDN is another 5G enabling technology which will helpto design dynamic manageable cost-effective and adaptablenetworks SDN has fuel the advancement of network soft-warization by proposing to transfer the control functionalityto software based entities ie network controllers SDN elim-inates the use of vendor specific black-box hardware therebypromoting the use of commodity servers and switches overproprietary appliances

Notwithstanding the transfer of network control function-alities to software based centralized entities demands thedata plane devices to communicate frequently with the SDNcontrollers Thus SDN controllers are located closer to thedata plane to reduce the latency in packet processing MECoffers the opportunity to locate control functions closer todata plane devices Moreover MEC complements the SDNadvancement of the transformation of the mobile-broadbandnetwork into a programmable world ensuring highly efficientnetwork operation and service delivery [167] Thus the popu-larity of SDN in different domains including 5G IoT will fuelthe adaption of MEC concept as well

Many recent research works justify the added benefits ofthe combine use of SDN and MEC in IoT systems [168]ndash[175] The role of NFV and SDN in MEC ecosystem isdiscussed in [168] SDN can be also used to make MECmore flexible and cost-effective for 5G applications The real-time heart attack mobile detection service proposed in [169]is a novel e-health IoT service that employs SDN-poweredMEC in a Vehicular Ad-hoc Network (VANET) architecturefor reliable performance In [170] a novel SDNNFV-basedsecurity framework is presented to enable integrated protectionfor IoT systems and in MEC applications An SDN-basedMEC framework has been proposed to provide the requireddata-plane flexibility programmability and reduced latency forapplications such as VR and Vehicular IoT [171]

In addition a conceptual approach to providing security forIoT systems by using SDN and edge computing is presentedin [176] The SDN-based IoT mobile edge cloud architecture(SIMECA) proposed in [172] can deploy diverse IoT servicesat the mobile edge by leveraging distributed lightweightcontrol and data planes optimized for IoT communicationsIn [173] the utilization of SDN and MEC to overcome thechallenges of network densification of IoTcloud integrationover a smart home is presented Likewise the MEC-SDNframework presented in [175] guarantees the QoS requirementsatisfaction and efficient use of the wireless resources intactical network applications

C Information Centric Networking

To address the ever increasing traffic volume in the Internetapplications such as HD mobile video ARVR 3D gamingand cloud computing a new set of network architecturesand networking technologies are developed over the past fewdecades These technologies employ caching replication andcontent distribution in optimum ways Among them ICNhas become one of the main approaches to addressing thisdemand [177] [178] ICN is an Internet architecture that putsinformation at the center where it needs to be and replacesthe client-server model by proposing a new publish-subscribemodel The key benefits of ICN include fast and efficient datadelivery and improved reliability Thus ICN is considered oneof the promising networking models for IoT ecosystem

MEC and ICN are complementary concepts which can bedeployed independently [67] However both could add valueto 5G and IoT domains in a complementary fashion Certainsynergies can be exploited when these two technologies aredeployed cooperatively For example ICN can be used for con-tent distribution over an unreliable radio links and transparentmobility among multiple technologies [179] while MEC canbe used to reduce the latency for delay critical applicationssuch as tactile Internet [180] and ARVR applications or toperform distributed data-reduction and security functions foran IoT network

In addition the use of MEC with ICN can further improvethe performance of edge computing It can solve some of theexisting challenges in MEC ecosystem For instance MECis facing a challenge of application level reconfigurationsince it requires a re-initialization of the session whenever

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

19

a session is being served by a non-optimal service instanceSuch application level reconfiguration will increase the delayin session migration However the natural support for service-centric networking in ICN can minimize the network relatedconfiguration for applications It will reduce the reconfigura-tion delay and allow fast resolution for named service instances[181]

ICN can also improve the edge storage and caching featuresof MEC enabled networks ICN allows location independentdata replication and opportunistic caching at strategic pointsin the network These features benefit both real-time and non-realtime IoT applications where a set of IoT devices or usersshare the same content [181]

Opportunities and challenges of MEC and ICN integrationfor IoT are presented in [182] Here the authors highlightthe synergies that can be exploited when the two technologiesare deployed cooperatively for IoT applications In additionseveral research works have also verified the importance ofICN and MEC cooperation [67] [183]ndash[186] A novel HetNetsvirtualization architecture with ICN and MEC techniques isproposed for video trans-coding caching and multi-cast in[183] A virtual multi-resources allocation scheme is used inthe designed framework to maximize the utility of computingcaching and communication to support the massive contentdelivery The vision of combining ICN and MEC in thecontext of connected vehicle environments is presented in[67] It shows how ICN in combination with MEC can addressthe challenges of futuristic vehicular application scenarios Anovel information-centric heterogeneous networks frameworkis proposed in [184] to optimize the virtual resource allocationat the edge Authors formulate the virtual resource allocationstrategy as a joint optimization problem by considering bothvirtualization and caching and computing at the edge A novelframework which jointly considers networking caching andcomputing techniques to support energy-efficient informationretrieval and computing services is presented in [185] Thisframework integrates SDN MEC and ICN to enable thedynamic orchestration of different resources in next generationgreen wireless networks A MEC-enabled ICN-based contenthandling framework at the mobile network edge is presented in[186] The proposed framework realizes context-aware contentlocalization in order to enhance user QoE in video distributionapplications

D Network Slicing

Network slicing proposes a way of separating the networkinto different network segments Thus it allows multiplelogical network segments to be created on top of a commonshared physical infrastructure [187] Future IoT will enablea wide range of different types of connections and servicesThese connections and services will need performance guar-antees as well as security Network slicing can satisfy theserequirements Moreover 5G mobile network will support bothMEC and network slicing technologies [188]

Network slicing can be used in different IoT domains Oneof such application domain is massive IoT [189] In orderto support massive IoT systems the network should be able

to satisfy requirements such as massive cost reduction incommunication network scalability and edge analytics Theintegration of MEC with Network slicing can be used tosatisfy some of these requirements such as scalability andedge analytics Another use case is critical communicationsfor delay critical applications such healthcare autonomousdriving and industrial Internet The key requirements to enablesuch critical communications are reduced latency and trafficprioritization While MEC can be used to reduce latencynetwork slicing can support traffic prioritization

Figure 5 illustrates the utilization of network slicing indifferent applications Here network slicing can be use todivide the MEC resources in to different slices dynamically Itwill improve the efficiency of using MEC resources in differentIoT applications

Fig 5 Use of Network Slicing in different applications [190]

Several research articles already presented the possibility ofusing Network slicing with MEC to provide improved servicesfor IoT and other 5G applications

An overview of the Third Generation PartnershipProject (3GPP) standard evolution from network sharingprinciples mechanisms and architectures to future on-demandmulti-tenant systems is presented in [187] MEC is identifiedas one of the key attributes to realize the aforementionednetwork slicing extensions in 3GPP toward full multi-tenancyA logical architecture for network-slicing-based 5G systemsis presented in [191] Here authors show the evolution ofnetwork slicing in network architecture and the synergy withSDN NFV and MEC technologies The work presented in[192] discusses the design challenges of network slicing withother concepts such as cloud-RAN and MEC A SDNNFVpacketoptical transport network and edgecore cloud platformfor E2E 5G and IoT services is presented in ADRENALINTEtestbed [193] It demonstrates the use of SDNNFV controlsystem to provide the global orchestration of the multi-layer(packetoptical) network resources and network slicing baseddistributed cloud infrastructure for multi-tenancy

Table XII summarizes the reviewed state-of-the-art MEC-IoT integration technologies

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

20

TABLE XII Comparison of the reviewed state-of-the-art MEC-IoT Integration Technologies

Ref Description IoT applicationOR domain

NFV SDN ICN NetworkSlicing

[164] Present a NFV-enabled MEC architecture for video streaming gamingand AR

Gaming and AR X

[163] Present an double-tier MEC-NFV integrated architecture for 5G appli-cations

Gaming X

[166] Present an integrated orchestration solution by combining the NFV andMEC use cases within a single orchestration environment

General IoT X

[165] Present a NFV-enabled MEC framework for low latency mobile appli-cations

General IoT X

[194] Preset an network architecture to addresses some of the central con-vergence challenges of NFV 5GMEC IoT and fog

General IoT X

[168] Discuss the role of NFV and SDN in MEC and IoT ecosystem General IoT X X[169] Present SDN-MEC based Real-time Heart Attack Mobile Detection

Service (RHAMDS) by using smart watchesehealth WIoT X

[170] Present SDN-NFV based security framework which can integrationwith existing IoT security mechanisms

General IoT X X

[171] Present SDN-based MEC framework for low latency applications VR and IoT Au-tomotives

X

[195] Present an SDNNFV architecture to delivery of future 5G servicesacross multiple technological and administrative networks

General IoT X X

[176] Present an conceptual approach to provide security for IoT systems byusing SDN and edge computing

General IoT X

[172] Presnet an SDN-based IoT Mobile Edge Cloud Architecture (SIMECA)for future IoT applications

General IoT X X

[174] Present a four-tier architecture assisted by MEC and SDN for VANETs IoT Automotive X[173] Discuss the utilization of SDN and MEC to overcome the challenges

of network densificationSmart homes X

[175] Present an MEC and SDN based framework for efficient and flexibleservice delivery

Tactile Internet X

[182] A white paper on opportunities and challenges of MEC and ICNintegration for IoT

General IoT X

[183] Present an novel HetNets virtualization architecture for video trans-coding caching and multi-cast

VRAR GamingWIoT

X

[67] Present the vision of combining ICN and MEC in the context ofconnected vehicle environments

IoT Automotive X

[184] Present a novel information-centric heterogeneous networks frameworkfor virtual resource allocation at the edge

General IoT X

[185] Present a novel framework which jointly considers networking cachingand computing techniques to support energy-efficient information re-trieval and computing services

General IoT X X

[186] Present a content handling framework which realizes context-awarecontent localization to enhance user QoE in video distribution applica-tions

VRAR GamingWIoT

X

[196] Propose an 5G-ICN architecture to realize an ICN-based servicedelivery for future IoT applications

General IoT X X X

[189] A discussion on use of network slicing for Massive IoT services General IoT X[192] Propose an novel network slicing architecture for integrated 5G com-

munications including IoTGeneral IoT X

[193] Propose an packetoptical transport network and edgecore cloud plat-form and testbed implementation for E2E 5G and IoT services

General IoT X X X

V PROJECTS

The European 5G PPP (5G Infrastructure Public PrivatePartnership) is one of the key layers on efforts to leverageMEC and IoT technologies to support the evolution towards5G networks In this section we discuss some renownedongoing EU research projects which are explicitly contributingto MEC and IoT technologies These projects along withtheir technological aspects and the key research areas aresummarized in Table XIII Since the concept of MEC wasinitiated by ETSI all of these projects are EU based Howeverthey have other non-EU partners as Japan Taiwan and ChinaThe recent Horizon 2020 (H2020) funding scheme has fueledthe MEC related research in Europe with the cooperation ofother parts of the globe Although non-EU international levelprojects are hardly found on integrating MEC and IoT the

other countries have projects on different edge technologiesincluding MCC fog and cloudlets We have excluded theseprojects from our survey since they are out of scope from themainstream of the paper

1) SESAME Small cEllS coordinAtion for Multi-tenancyand Edge services (June 2015 - Dec 2017) SESAME [197] isone of the front-line EU H2020 projects which focuses on thedevelopment and demonstration of an innovative architecturecapable of providing Small Cell (SC) coverage to multiplevirtual operators as-a-Service This is a pioneering project thatuses MEC and NFV technologies to realize the cloud-enabledsmall cell (CESC) concept by supporting powerful self-x (xstands for organizing optimizing or healing) managementfeatures and executing novel applications and services insidethe access network infrastructure SESAME is expected to

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

21

deliver the small cell concept in high dense 5G scenariosMoreover it intends to consolidate multi-tenancy in communi-cations infrastructures This allows several operators or serviceproviders to engage in new sharing models of both accesscapacity and edge computing capabilities

2) ANASTACIA Advanced Networked Agents for Securityand Trust Assessment in CPS IOT Architectures (Jan 2017 -Dec 2019) ANASTACIA [198] an EU H2020 funded projectwhich promises to develop and demonstrate a holistic solu-tion enabling trust and security by-design for heterogeneousdistributed and dynamically evolving CPS based on IoT andvirtualised cloud architectures The security framework withself-protection self-healing and self-repair capabilities willbe designed in full compliance to SDNNFV standards Thiswill include the security development paradigm distributedtrust and security enabler and dynamic security and privacyseal In particular ANASTACIA will address the securitychallenges in two use cases on the deployment of MEC serverand smart buildings

3) 5G-MiEdge Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystem (July 2016 - June 2019) 5G-MiEdge [199]is a publicly supported research project bringing Millimeter-Wave (mmWave) technology and MEC into the mobile radioworld It was co-funded by EU H2020 and Japanese gov-ernment It combines mmW accessbackhauling with MECto enable enhanced mobile broadband (eMBB) services andmission critical low-latency applications using cost-efficientRANs The project is composed of three key technologiesnaming the protocols of mmWave accessbackhaul links ultra-lean and inter-operable control signaling mechanism (liquidRAN C-plane) over 3GPP LTE and user or application centricorchestration algorithms for edge resource allocation 5G-MiEdge intends to develop transmission schemes and pro-tocols of mmWave accessbackhauling which can assist themobile edge cloud with cachingprefetching This will beuseful in realizing ultra-high speed and low latency servicedelivery which will be resilient to network bottlenecks such asbackhaul congestion usersrsquo density and mission-critical ser-vice deployments The targeted use cases are mostly stadiumsoffices and train stations

4) 5GPagoda 5GPagoda project [200] aims at creating avirtual mobile network that can be deployed upon requestdedicated to an application to be used during the TokyoOlympic Games in 2020 5GPagoda intends to develop ascalable 5G slicing architecture and a highly programmablenetwork control and data path supporting mechanism foruse cases in IoT and human communication This would beachievable through the development of a scalable network slicemanagement and orchestration frameworks These frameworkswould serve distributed edge dominated network infrastruc-tures and convergent software functionality for lightweightcontrol plane and data plane programmability

5) Inter-IoT (Jan 2016 - Dec 2018) Horizon 2020 EUproject INTER-IoT project [201] aims to design implementand test an open framework that will allow interoperabilityamong different IoT platforms The project uses a layer-oriented approach for the interoperability framework in fourapplication domains smart grid e-health smart factories and

transport-logistics The final goal is to integrate different IoTdevices networks platforms services and applications thatwill allow a global continuum of data infrastructures andservices which can enable different IoT use cases

6) 5G-MoNArch 5G Mobile Network Architecture for di-verse services use cases and applications in 5G and beyond(July 2017 - June 2019) 5G-MoNArch [202] is anotherproject funded by EU Horizon 2020 programme and it willevolve 5G-PPP Phase 1 concepts to a fully-fledged archi-tecture develop prototype implementations and apply theseprototypes to representative use cases 5G-MoNArchs specifictechnical goal is to use network slicing which capitalizeson the capabilities of SDN NFV orchestration of accessnetwork and core network functions and analytics to supporta variety of use cases in vertical industries such as automotivehealthcare and media The devised 5G-MoNArch architecturewill be deployed in two test beds a sea port and a tourist city

7) 5G-ESSENSE Embedded Network Services for 5G Ex-periences (June 2017 - June 2019) 5G ESSENCE [203] is anEU H2020 funded project that proposes a highly flexible andscalable 5G small cell platform leveraging the paradigms ofedge cloud computing and Small-Cell-as-a-Service ESSENCEbuilds virtualization techniques on the distributed and network-integrated cloud inherited by 5G-PPP Phase 1 SESAMEproject that provides processing power at the edge of thenetwork The project will explicitly address two use casesincluding in-flight entertainment and connectivity systems andmission critical applications for public safety

8) MATILDA (June 2017 - June 2019) The EU H2020funded 5G-PPP Phase 2 project MATILDA [204] aims todesign and implement a holistic 5G framework for the designdevelopment and orchestration of 5G-ready applications and5G network services over a sliced programmable infrastruc-ture using VNFs Intelligent and unified orchestration mecha-nisms will be applied for the automated placement of the 5G-ready applications and the creation and maintenance of therequired network slices The management of the cloudedgecomputing and IoT resources is supported by a multi-sitevirtualized infrastructure manager

9) 5GCITY (June 2017 - June 2019) 5GCity [205] is alsoan EU H2020 funded 5G-PPP Phase 2 project which demon-strates how to empower the city infrastructure and transformthem into a hyper-connected distributed 5G-enabled edgevirtualization domain The project targets three different cities(Barcelona Bristol and Lucca) and would benefit telecommu-nication infrastructure providers municipalities and a numberof different vertical sectors utilizing the city infrastructure Itwill leverage the virtualization platform in order to enable thecities to create dynamic E2E slices containing both virtualizededge and network resources and lease to third-party operators

10) MONICA Management Of Networked IoT WearablesVery Large Scale Demonstration of Cultural and SocietalApplications (Jan 2017 - Dec 2019) MONICA [206] is an EUH2020 funded large scale pilot project which aims to providea very large scale demonstration of multiple existing and newIoT technologies for smarter living It demonstrates a largescale IoT ecosystem that uses innovative wearable and portableIoT sensors and actuators with closed-loop back-end services

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

22

integrated into an interoperable cloud-based platform capableof offering a multitude of simultaneous targeted applicationsThe key objectives of this project are to strengthen crowdsafety and security at at big cultural open-air events andimprove user experience Given these goals the final solutionshould be compatible with many different IoT sensors opensource with cost effective wearables and strengthened withdata security privacy and trust

11) AUTOPILOT AUTOmated driving Progressed by Inter-net Of Things (Jan 2017 - Dec 2019) Another large scale pilotproject funded by EU H2020 AUTOPILOT [207] will deploytest and demonstrate IoT-based automated driving use casescomprising urban driving highway pilot automated valetparking and platooning The project will integrate into vehicleIoT sensors and use cloud and MEC type IoT platforms (egBrainport pilot site in Netherlands) to share sensor data andcreate new autonomous mobility services The AUTOPILOTproject will create and deploy new business products andservices for fully automated driving vehicles used at the pilotsites This project will feature innovations such as driving routeoptimization vulnerable road user sensing and dynamicallyupdating an IoT based HD map

12) 5G-CORAL A 5G Convergent Virtualised Radio AccessNetwork Living at the Edge (Sep 2017 - Aug 2019) The newly initiated EU H2020 project 5G-CORAL [208]leverages on the pervasiveness of edge and fog computing inRAN to create a unique opportunity for access convergenceThis is envisioned by the means of virtualised networking andcomputing solution where virtualised functions context-awareservices and user and third-party applications are blendedtogether to offer enhanced connectivity and better qualityof experience The proposed solution considers two majorbuilding blocks namely the edge and fog computing systemand the orchestration and control system 5G-CORAL projectwill be validated in three testbeds a shopping mall high-speedtrain and connected cars

VI LESSONS LEARNED AND FUTURE RESEARCHDIRECTIONS

In this section we present the lessons learned and thefuture research directions with respect to MEC-IoT integrationIn particular we focus on MEC-IoT application paradigmstechnical aspects (ie scalability communication computa-tion offloading and resource allocation mobility managementsecurity privacy and trust management) and standardizationefforts

A Applications

1) Lessons learned MEC is an ideal solution that supportsthe increased demand for bandwidth consumption and ultralow latency requirements of IoT applications MEC resourcescan be utilized for the pre-processing of massive IoT datawhich will reduce bandwidth consumption provide networkscalability and ensure a fast response to user requests How-ever in order to reap the maximum benefits of MEC for IoTthere needs to be more in dept research on how to efficientlydistribute and manage data storage and computing resources

at the network edge Since MEC is still not well establishedthere can be myriad of technical challenges that need tobe addressed Moreover due to much unprecedented userexpectations the requirements for designing MEC systemsmay vary upon the IoT application area

2) Future research directions The applications describedin Section II are overlapping in several ways For instanceAR and VR may explicitly support autonomous driving byexchanging information derived from multi-resolution mapscreated using the local sensors of the vehicles This willextend the visibility of the vehicle The edge servers areexpected to perform pro-actively in such AR and VR systemsTele-surgery is another domain that takes advantage of ARand VR exploitation In the ideal situation VR should haveno distinction between real and virtual worlds In order toachieve this goal the concepts of MEC in VR applicationsmight be merged with concepts like quantum computing Itis reported that ETSI and Virtual RealityAugmented RealityAssociation (VRARA) intend to collaborate on interactive VRand AR technologies delivered over emerging 5G networksand hosted on MEC sites [209] VRARA will encouragecommon member companies to pursue VRAR-focused usecases and requirements for ETSI MEC Phase 2

The adoption of machine learning techniques in 5G net-works has increasingly attracted the attention of the researchcommunity This will provide adaptive learning and decision-making approaches to meet the requirements of different verti-cals The integration of Artificial Intelligence (AI) algorithmsand machine learning at the edge of the networks will furtherassist the data-intensive requirements of the IoT applicationsParticularly AI techniques can be exploited for adaptiveoptimal and pro-active action on instantaneous networkingdemand in vehicular communications in the context of self-driving vehicles However more efforts are needed to adoptmachine learning techniques such as recursive neural net-works reservoir computing and deep learning in autonomousvehicles kind of applications due to their complex networkarchitecture and enormous data sets More importantly thereis no unifying theories to define how such a network willbehave

B Scalability

1) Lessons learned Several aspects of the present-dayscalability schemes and data management paradigms will needsubstantial refinement in order to be able to handle the changesthat are expected in future MEC-enabled IoT networks IoTdevices like sensors and RFID capturing devices are expectedto keep capturing objects almost in real-time hence generatinga huge amount of readings Timeliness is another factor in suchscenarios since generated data usually have very short life-span of about 2 seconds Obviously the present-day approachto information search and data management cannot handlethis expectation in a scalable manner For this reason a morerefined search and indexing algorithm will be required for bothMEC-enabled IoT applications and IoT systems in general

2) Future research directions The adoption of the IPv6is a significant move that will further advance scalability

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

23

TABLE XIII Contribution of global level ongoing projects on MEC and IoT Todo Shall we remove mmWave here We didnot discuss that a lot

Project

SESA

ME

[197

]

AN

AST

AC

IA[1

98]

5GM

i-E

dge

[199

]

5GP

agod

a[2

00]

Inte

r-Io

T[2

01]

5G-M

oNA

rch

[202

]

5G-E

SSE

NSE

[203

]

MA

TIL

DA

[204

]

5GC

ITY

[205

]

MO

NIC

A[2

06]

AU

TOPI

LO

T[2

07]

5G-C

OR

AL

[208

]

TechnologiesMEC X X X X X X X X X X XIoT X X X X X X X X X X XSDN X X X X X X X XNFV X X X X X X X XNetwork slicing X X X X X XmmWave X

Research focusNetwork architecture OR framework X X X X X X X X X XCommunication and network infrastructure X X X X XComputation offloading X XResource management X X X X XMobility X X XScalability X X XInteroperability X XSecurity X X X XPrivacy X X XTrust X X

in MEC-enabled IoT applications going forward In [210]authors proposed the idea of CONCERT a term coined fromthe combination of cloud and cellular system The CONCERTsolution exploits the principles of NFV and SDN to enhancescalability in future networks Since scalability is a hugefactor to determine where the MEC server gets deployedand since the devices exploiting the MEC server located inthe core network will inevitably experience longer latenciesthen there could be a major hindrance to the use of real-timeapplications in such MEC settings Regarding control signalingin MEC the proposed CONCERT approach also adopts eithera fully centralized control or a hierarchical control for betterscalability and flexibility

C Communication1) Lessons learned As MEC is still at its infancy defining

a solid communication model for the entire MEC architectureis an open research question that paves many opportunitiesto the academia industry and the standardization entitiesAdvanced wireless communication techniques are required todesign for interference cancellation and adaptive power controlat the MEC servers in order to reduce the offloading energyconsumption in a significant manner The tight alliance be-tween MEC and IoT may also create new research challengesin communication perspective

2) Future research directions As pointed out by Raza etal in [15] interoperability among various IoT LPWAN tech-nologies encountered in IoT is still an open research questionto address There are still insufficient testbeds and open-sourcetool chains for LPWAN technologies Massive connectivityand high data rate requirements of IoT devices (eg wear-ables) can be fulfilled by accompanying new radio access tech-nologies such as Non-Orthogonal Multiple Access (NOMA)and massive Multiple-Input-Multiple-Output (MIMO) [76]

Moreover many research efforts on edge caching are under-way to achieve the trade-off between the transmission rate andstorage at the MEC hosts [107] The co-existence of differentwireless communication technologies available for IoT maystill create many challenges for edge level accessibility sincethe IoT applications are diversified in versatile areas whereeach has a unique set of requirements Furthermore they haveconflicting goals such as energy efficiency high throughputand wide coverage Therefore system-level research is re-quired to reap out the maximum benefit on exploiting suchcommunication technologies

Implementing MEC over FiWi access networks are inves-tigated due to their low costs wide deployments and highcapacity [110] These fiber-wireless broadband access net-works may provide a single communication platform for MECand centralized cloud services over the wired and wirelessnetworking technologies ICN in combination with MEC isidentified as another promising way of establishing a com-munication model for vehicular networks [67] where movingvehicles may incur frequent disconnects and re-connects todifferent network access points

D Computation Offloading and Resource Allocation

1) Lessons learned Decision making for data offloadingat the user-end devices and the resource allocation for thoseoffloaded dataapplication at the edge clouds are two highlyregarded topics discussed among the research communityespecially those who engaged in MEC and IoT eras Mostof the prior works were focused on the offloading mecha-nisms for latency critical applications while minimizing energyconsumption at the UE In contrary IoT permits a platformthat has both delay sensitive and delay tolerant applicationsAlthough most of the proposed solutions are evaluated by

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

24

means of theoretical analysis or simulations there is still noproper formation of standard offloading mechanism for IoTand MEC systems

2) Future research directions Mobility is a principal fea-ture of IoT devices which are either being transported byhumans (eg wearable sensor) or by another carrier (egvehicular networks) or being mobile by itself (eg robots)Mobility-aware resource management and computation of-floading strategies need to be precisely investigated in theera of IoT supportive MEC systems Scalability is the otherequally important feature to consider in large scale IoT de-ployments where edge computing needs seamless offloadingand resource allocation policies Other accelerating tendenciestowards future research efforts in the field of MEC and IoTmay include server cooperation in MEC dependency-awareoffloading and dynamic resource allocation

The exploitation of Knowledge-Defined Networking (KDN)to make intelligent predictions about offload costs can be lever-aged for efficient resource allocation at MEC servers as well asthe offloading decision making at IoT devices [211] The newparadigm of KDN is composed of Network Analytics (NA)SDN and AI techniques The introductory work in [212] pro-poses an intelligent computation offloading framework basedon user dynamics and historical data

E Mobility Management

1) Lessons learned Mobility management in MEC-enabled IoT has attracted a lot of attention in both researchand the industry This comes natural given that mobile nodesare expected to dominate the future IoT networks An optimaloffloading decision will be necessary for effective integrationof MEC with IoT Thus far most of the works on mobilitymanagement in the context of MEC are solely focusing onoptimizing the energy consumption at IoT nodes Howeverdesigning efficient and optimal MEC-enabled IoT systems willrequire energy optimization at the MEC end also This includesenergy consumed on computation and energy consumed oncommunication

Furthermore most works on offloading decisions are basedon static scenarios where the IoT device moves from one MECeNB to another and remains in one steady location during theoffload which is not necessarily the situation in most cases

2) Future research directions The energy required foroffloading or handover could vary substantially based on themovement factor during the offload [139] For this reasonthere will be a need for more advanced decision making al-gorithms They will leverage on various prediction techniquesto determine when offloading is in fact necessary what thechannel quality will be like during the offloading and whatthe entire offloading process will cost for each offloadingcondition

For advancing the VM migration techniques a crucial stepmoving forward is to optimize the migration process byminimizing the time required to complete a full migrationThis will mostly dependent on the protocol design of themigration process Hence an optimal solution is requiredfor a collaborative effort on the side of individuals and

organizations That notwithstanding still the VM migrationscheme might not be suitable for highly delay-sensitive real-time applications In general to achieve an efficient and highlyoptimized mobility management scheme for MEC-enabled IoTapplications there will be a need for a more holistic approachSuch a solution will encompass power control VM migrationdata compression and path selection [29]

F Security

1) Lessons learned Notwithstanding the closed paradigmof MEC it is important to realize that the whole ecosystemof MEC will not be controlled by one single owner orservice provider MEC data centers are capable of providingservices without relying on centralized infrastructures Thusit is certain that all MEC relevant assets such as the networkinfrastructure the service infrastructure (eg edge data centerscore infrastructure) the virtualization infrastructure and theuser devices will not be controlled by a single entity Thescale of this effect is further confounded by the diversitythat exists in IoT applications Consequently every element ofMEC and IoT infrastructure should be targeted towards globalnetworking environment As discussed in [145] the ldquoanythinganytimerdquo principle should be the underlying building blocksand application scenarios for MEC-enabled IoT systems [154]Conversely the ldquoanywhererdquo principle also implies that attackscan be performed from anywhere making the edge paradigmsa double-edged sword and hence the need for security mea-sures that span the entire global networking paraphernalia

2) Future research directions The future of MEC-enabledIoT systems will revolve mostly around developing universalstandard security mechanisms that can adequately protectthe whole ecosystem against security threats Such universalstandards will enable both service providers and developers tounderstand the particularities of every edge paradigm as theyhave subtle differences that will affect the implementation anddeployment of the security mechanisms [145] Currently theabsence of such global perimeters is seen as one of the banesto the security of the edge paradigms

One notable effect of the lack of a global perimeter is thenature of the different attacker profiles that will target edgeparadigms [213] In the present day networks adversaries aremostly external entities with no stake in control of networkelements However with the advent of MEC-enabled IoT thereexist many adversaries that will control one or more elementsof the infrastructure such as user devices VMs serverssections of the network and in the worst case an entire edgedata center [154] Adopting deep-learning-based models at theedge level to detect malicious applications will be anotherinteresting research area Applying reinforcement learningtechniques to develop edge security solutions can be exploitedfor anomaly detection and lightweight authentication

G Privacy

1) Lessons learned The rise of new architecture newtechnologies and new network services will open up newchallenges to privacy protection On the one hand the existingprivacy objectives are outdated and are not compatible with

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

25

current technologies such as MEC IoT and 5G Thereforethese privacy directives have to be updated Governing orga-nizations have already started redefining the privacy objectivesFor instance the European commission adopted a GeneralData Protection Regulation (GDPR) in April 2016 It willbe superseded by the data protection directive and is plannedto be enforceable starting on 25 May 2018 On the otherhand privacy awareness is significantly increasing among thegeneral public users [214] Therefore the future networksrequire to provide an extra level of privacy than the earliergeneration of networks

2) Future research directions The future research workshould be focused on addressing above privacy challengesNew privacy protection mechanisms such as Software DefinedPrivacy (SDP) [155] Privacy by Design (PbD) [215] and SDNbased privacy-aware routing [216] can be used to providethe required level of privacy while or after the integration ofMEC to IoT systems SDP [155] allows easy orchestrationsof existing tools for enforcing privacy requirements of anInfrastructure as a Service (IaaS) cloud customer This conceptcan further be extended to provide privacy protection for MECenabled IoT systems PbD is an approach in system engi-neering which promotes the integration of privacy throughoutthe whole design process [215] PbD approach can be usedduring the MEC integration in IoT systems If SDN is usedin MEC-IoT systems which is highly likely user data packetscontaining privacy information that should not cross localspaces or even country borders could be identified Thenthe SDN controller could define flow rules so that thesepackets are routed only via the links and routers with highsecurity More sophisticated routing protocols can be designedby increasing the number of such qualifiers

H Trust Management

1) Lessons learned Trust management in MEC systemsis still a barely investigated area In order to strengthen theuser ecosystem in centralized cloud environment a flexibletrust manager can be shared among the cloud infrastructureproviders [217] Likewise the mutual trust should be incorpo-rated among the MEC servers to enhance the secure sharingof IoT datasets

2) Future research directions Context-aware trust relation-ships based on social computing are yet to be investigated inthe paradigm of IoT and edge computing A comprehensivetrust framework is still lacking for holistic trust managementin IoT with the context of MEC which is capable of achievingall the objectives listed above and fulfills the requirementsfrom different trust levels Future research needs to focuson data collecting at IoT perception layer and processing atedge servers in order to improve the IoT and MEC servicequality Complex and resource consuming trust managementalgorithms are not affordable by the tiny IoT devices Further-more device and network heterogeneity in IoT raises furtherchallenges There are also some open research trends formaking light-weight trust management mechanisms suitablefor heterogeneous IoT

I Standardization

The standardization of the MEC technology is relativelyrecent and currently ongoing The goal is to bring togetherall experts and industry players in consensus to define thecharacteristics and rules that will govern the implementationand interconnection of the MEC technology globally Justlike other standardized technologies the standardization ofMEC will open up an infinite avenue for developers andinnovators to harness the benefits of MEC in designing cutting-edge technologies and innovative solutions that will drive5G and future networks On the side of the customers suchstandardization would by no small measure affirm their trustin MEC and other related products and services

1) Future research directions The standardization pro-cesses of MEC along with the coordination and managementtasks are lead by an ETSI ISG [71] The MEC ISG groupaims at creating an open standardized and efficient platformfor the seamless integration of enterprise applications fromdifferent vendors and service providers into the MEC platformMost recently the 3GPP has shown a growing interest inincorporating MEC into its 5G standard and has identifiedfunctionality supports for edge computing in a recent technicalspecification contribution

The standardization entities are required to ensure thatMEC architecture works harmoniously with the heterogeneousIoT echo systems and related technologies Moreover sincethere are numerous third-party partners such as applicationdevelopers content providers and network device vendors thecomplexity of the services and the management of very largescale environment becomes challenging [218]

It is also important to do security and privacy legislationand standardization in a global context Different jurisdictionsshould cooperate together to develop inter-operable securityand privacy requirements to facilitate the flow of informationwith the required level of protection Thus the security andprivacy regulations will play a vital role to promote the adapta-tions new technologies such as MEC Regulatory entities suchas governments and standardization organizations have to worktogether with industry to define andor update the regulationsaccording to the new technologies

VII CONCLUSIONS

The advancements of MEC and IoT technologies will becontributing immensely to the realization of the highly an-ticipated game-changing vision of 5G and future generationsof mobile networks The propounders of MEC which isrelatively a recent technology have identified IoT as one ofthe important use cases of MEC MEC server performs as agateway between the latency critical and massive IoT networksand the core network where it can provide edge-cloud comput-ing and networking functionalities IoT application domainsare empowered with MEC technology by extending someintelligence to the edge of the network Although MEC willprovide on-site cloud computing services for IoT networksthere are still challenges in terms of device and networkheterogeneity scalability mobility and security In additionto the possible future works discussed in Section VI there are

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

26

few other research topics including but not limited to MECservice level congestion control latency aware routing anddynamic application routing In all essence MEC and IoT aretwo complementary technologies that if well harnessed havethe potential of advancing the course of the 5G networks andbeyond

ACKNOWLEDGMENT

This work has been performed under the framework of theInfotech Doctoral Program of UniOGS and the three projects6Genesis Flagship (grant 318927) SECUREConnect (SecureConnectivity of Future Cyber-Physical Systems) and TowardsDigital Paradise This research is funded by Academy ofFinland and TEKES Finland

REFERENCES

[1] P Guillemin and P Friess ldquoThe Industrial Internet of ThingsVolume G1 Reference Architecturerdquo The Cluster of EuropeanResearch Projects Tech Rep September 2009 [Online]Available httpwwwinternet-of-things-researcheupdfIoT ClusterStrategic Research Agenda 2009pdf

[2] Y-C Hu M Patel D Sabella N Sprecher and V Young ldquoMobileEdge ComputingA Key Technology Towards 5Grdquo ETSI White Papervol 11 no 11 pp 1ndash16 2015

[3] P Schulz M Matthe H Klessig M Simsek G Fettweis J AnsariS A Ashraf B Almeroth J Voigt I Riedel et al ldquoLatency CriticalIoT Applications in 5G Perspective on the Design of Radio Interfaceand Network Architecturerdquo IEEE Communications Magazine vol 55no 2 pp 70ndash78 2017

[4] T Taleb K Samdanis B Mada H Flinck S Dutta and D SabellaldquoOn Multi-Access Edge Computing A Survey of the Emerging 5GNetwork Edge Architecture amp Orchestrationrdquo Communications Surveysamp Tutorials vol 19 no 3 pp 1657ndash1681 2017

[5] L Atzori A Iera and G Morabito ldquoThe Internet of Things A SurveyrdquoComputer networks vol 54 no 15 pp 2787ndash2805 2010

[6] J Gubbi R Buyya S Marusic and M Palaniswami ldquoInternet ofThings (IoT) A Vision Architectural Elements and Future DirectionsrdquoFuture generation computer systems vol 29 no 7 pp 1645ndash16602013

[7] A Al-Fuqaha M Guizani M Mohammadi M Aledhari andM Ayyash ldquoInternet of Things A Survey on Enabling TechnologiesProtocols and Applicationsrdquo IEEE Communications Surveys amp Tuto-rials vol 17 no 4 pp 2347ndash2376 2015

[8] V Gazis ldquoA Survey of Standards for Machine-to-Machine and the In-ternet of Thingsrdquo IEEE Communications Surveys amp Tutorials vol 19no 1 pp 482ndash511 2017

[9] M Weyrich and C Ebert ldquoReference Architectures for the Internet ofThingsrdquo IEEE Software vol 33 no 1 pp 112ndash116 2016

[10] S Sicari A Rizzardi L A Grieco and A Coen-Porisini ldquoSecurityPrivacy and Trust in Internet of Things The Road Aheadrdquo ComputerNetworks vol 76 pp 146ndash164 2015

[11] J Granjal E Monteiro and J S Silva ldquoSecurity for the Internet ofThings A Survey of Existing Protocols and Open Research IssuesrdquoIEEE Communications Surveys amp Tutorials vol 17 no 3 pp 1294ndash1312 2015

[12] P Porambage M Ylianttila C Schmitt P Kumar A Gurtov andA V Vasilakos ldquoThe Quest for Privacy in the Internet of ThingsrdquoCloud Computing vol 3 no 2 pp 36ndash45 2016

[13] Z Yan P Zhang and A V Vasilakos ldquoA Survey on Trust Managementfor Internet of Thingsrdquo Journal of network and computer applicationsvol 42 pp 120ndash134 2014

[14] L Atzori A Iera G Morabito and M Nitti ldquoThe Social Internet ofThings (SIoT)ndashwhen social networks meet the internet of things Con-cept architecture and network characterizationrdquo Computer networksvol 56 no 16 pp 3594ndash3608 2012

[15] U Raza P Kulkarni and M Sooriyabandara ldquoLow Power WideArea Networks An Overviewrdquo Communications Surveys amp Tutorialsvol 19 no 2 pp 855ndash873 2017

[16] C Perera A Zaslavsky P Christen and D Georgakopoulos ldquoContextAware Computing for the Internet of Things A Surveyrdquo Communica-tions Surveys amp Tutorials vol 16 no 1 pp 414ndash454 2014

[17] J A Stankovic ldquoResearch Directions for the Internet of ThingsrdquoInternet of Things Journal vol 1 no 1 pp 3ndash9 2014

[18] J Lin W Yu N Zhang X Yang H Zhang and W Zhao ldquoA Surveyon Internet of Things Architecture Enabling Technologies Securityand Privacy and Applicationsrdquo Internet of Things Journal 2017

[19] D Miorandi S Sicari F De Pellegrini and I Chlamtac ldquoInternetof Things Vision Applications and Research Challengesrdquo Ad HocNetworks vol 10 no 7 pp 1497ndash1516 2012

[20] L Atzori A Iera and G Morabito ldquoUnderstanding the Internet ofThings Definition Potentials and Societal Role of a Fast EvolvingParadigmrdquo Ad Hoc Networks vol 56 pp 122ndash140 2017

[21] D Sabella A Vaillant P Kuure U Rauschenbach and F GiustldquoMobile-edge Computing Architecture The Role of MEC in theInternet of Thingsrdquo Consumer Electronics Magazine vol 5 no 4pp 84ndash91 2016

[22] P Garcia Lopez A Montresor D Epema A Datta T HigashinoA Iamnitchi M Barcellos P Felber and E Riviere ldquoEdge-centriccomputing Vision and challengesrdquo SIGCOMM Computer Communi-cation Review vol 45 no 5 pp 37ndash42 2015

[23] S Shahzadi M Iqbal T Dagiuklas and Z U Qayyum ldquoMulti-accessEdge Computing Open Issues Challenges and Future PerspectivesrdquoJournal of Cloud Computing vol 6 no 1 p 30 2017

[24] E Ahmed and M H Rehmani ldquoMobile Edge Computing Opportuni-ties Solutions and Challengesrdquo Future Generation Computer Systems2017

[25] Y Ai M Peng and K Zhang ldquoEdge Cloud Computing Technolo-gies for Internet of Things A Primerrdquo Digital Communications andNetworks 2017

[26] N Abbas Y Zhang A Taherkordi and T Skeie ldquoMobile EdgeComputing A Surveyrdquo IEEE Internet of Things Journal vol 5 no 1pp 450ndash465 2018

[27] A Ahmed and E Ahmed ldquoA Survey on Mobile Edge Computingrdquo in10th IEEE International Conference on Intelligent Systems and Control(ISCO) 2016 pp 1ndash8

[28] M T Beck M Werner S Feld and T Schimper ldquoMobile Edge Com-puting A Taxonomyrdquo in Proc of the Sixth International Conferenceon Advances in Future Internet 2014 pp 48ndash55

[29] P Mach and Z Becvar ldquoMobile Edge Computing A Survey on Archi-tecture and Computation Offloadingrdquo IEEE Communications Surveysamp Tutorials vol 19 no 3 pp 1628ndash1656 2017

[30] A C Baktir A Ozgovde and C Ersoy ldquoHow Can Edge ComputingBenefit from Software-Defined Networking A Survey Use Cases ampFuture Directionsrdquo Communications Surveys amp Tutorials vol 19no 4 pp 2359ndash2391 2017

[31] I Afolabi T Taleb K Samdanis A Ksentini and H Flinck ldquoNetworkSlicing amp Softwarization A Survey on Principles Enabling Technolo-gies amp Solutionsrdquo IEEE Communications Surveys amp Tutorials 2018

[32] Y Mao C You J Zhang K Huang and K B Letaief ldquoA Survey onMobile Edge Computing The Communication Perspectiverdquo Commu-nications Surveys amp Tutorials vol 19 no 4 pp 2322ndash2358 2017

[33] S Wang X Zhang Y Zhang L Wang J Yang and W WangldquoA Survey on Mobile Edge Networks Convergence of ComputingCaching and Communicationsrdquo Access vol 5 pp 6757ndash6779 2017

[34] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems vol 78 pp 680ndash698 2018

[35] M Perez S Xu S Chauhan A Tanaka K Simpson H Abdul-Muhsin and R Smith ldquoImpact of Delay on Telesurgical PerformanceStudy on the Robotic Simulator dV-Trainerrdquo International journal ofcomputer assisted radiology and surgery vol 11 no 4 pp 581ndash5872016

[36] ldquoUnlocking Commercial Opportunities From 4G Evolution to5Grdquo GSMA Network Tech Report accessed on 21032018[Online] Available httpswwwgsmacomfuturenetworkswpcontentuploads201602704 GSMA unlocking comm opp report v5pdf

[37] ldquoThe Business Case for MEC in Retail A TCOAnalysis and its Implications in the 5G Erardquo Inteltechnical White paper June 2017 accessed on 14032018[Online] Available httpsbuildersintelcomdocsnetworkbuildersthe-business-case-for-mec-in-retail-a-tco-analysis-and-its-implications-in-the-5g-erapdf

[38] ldquoPutting Sensors to Work in the Factory Environ-ment Data to Information to Wisdomrdquo accessed on29042018 [Online] Available httpsitpeernetworkintelcomputting-sensors-to-work-in-the-factory-environment

27

[39] B L R Stojkoska and K V Trivodaliev ldquoA Review of Internet ofThings for Smart Home Challenges and Solutionsrdquo Journal of CleanerProduction vol 140 pp 1454ndash1464 2017

[40] C Vallati A Virdis E Mingozzi and G Stea ldquoMobile-Edge Comput-ing Come Home Connecting Things in Future Smart Homes Using LTEDevice-to-Device Communicationsrdquo Consumer Electronics Magazinevol 5 no 4 pp 77ndash83 2016

[41] R Morabito R Petrolo V Loscrı and N Mitton ldquoEnabling aLightweight Edge Gateway-as-a-Service for the Internet of Thingsrdquoin 7th International Conference on the Network of the Future (NOF)IEEE 2016 pp 1ndash5

[42] X Sun and N Ansari ldquoEdgeiot Mobile Edge Computing for theInternet of Thingsrdquo Communications Magazine vol 54 no 12 pp22ndash29 2016

[43] K-K Nguyen and M Cheriet ldquoVirtual edge-based smart communitynetwork managementrdquo Internet Computing vol 20 no 6 pp 32ndash412016

[44] T Taleb S Dutta A Ksentini M Iqbal and H Flinck ldquoMobileEdge Computing Potential in Making Cities Smarterrdquo CommunicationsMagazine vol 55 no 3 pp 38ndash43 2017

[45] W Shi J Cao Q Zhang Y Li and L Xu ldquoEdge Computing Visionand Challengesrdquo Internet of Things Journal vol 3 no 5 pp 637ndash6462016

[46] M S Hossain and G Muhammad ldquoCloud-assisted Industrial Internetof Things (IIoT)ndashEnabled Framework for Health Monitoringrdquo Com-puter Networks vol 101 pp 192ndash202 2016

[47] S R Islam D Kwak M H Kabir M Hossain and K-S Kwak ldquoTheInternet of Things for Health Care A Comprehensive Surveyrdquo Accessvol 3 pp 678ndash708 2015

[48] W Shi and S Dustdar ldquoThe Promise of Edge Computingrdquo Computervol 49 no 5 pp 78ndash81 2016

[49] T X Tran A Hajisami P Pandey and D Pompili ldquoCollaborativeMobile Edge Computing in 5G Networks New Paradigms Scenariosand Challengesrdquo Communications Magazine vol 55 no 4 pp 54ndash612017

[50] D Singh G Tripathi A M Alberti and A Jara ldquoSemantic EdgeComputing and IoT Architecture for Military Health Services inBattlefieldrdquo in 14th Annual Consumer Communications NetworkingConference (CCNC) IEEE 2017 pp 185ndash190

[51] S Nunna A Kousaridas M Ibrahim M Dillinger C ThuemmlerH Feussner and A Schneider ldquoEnabling Real-time Context-awareCollaboration Through 5G and Mobile Edge Computingrdquo in 12thInternational Conference on Information Technology-New Generations(ITNG) IEEE 2015 pp 601ndash605

[52] S K Sharma and X Wang ldquoLive Data Analytics With CollaborativeEdge and Cloud Processing in Wireless IoT Networksrdquo Access vol 5pp 4621ndash4635 2017

[53] A M Rahmani T N Gia B Negash A Anzanpour I AzimiM Jiang and P Liljeberg ldquoExploiting Smart e-Health Gateways at theEdge of Healthcare Internet of Things A Fog Computing ApproachrdquoFuture Generation Computer Systems vol 78 pp 641ndash658 2018

[54] ldquo5G Security Making the Right Choice to Match your NeedsrdquoSIMalliance 5GWG technical White paper Feb 2016 accessed on12022018 [Online] Available httpsimallianceorg

[55] O Zakaria J Britt and H Forood ldquoInternet of Things (IoT) Automo-tive Device System and Methodrdquo Jul 25 2017 uS Patent 9717012

[56] W Balid H Tafish and H H Refai ldquoIntelligent Vehicle Counting andClassification Sensor for Real-Time Traffic Surveillancerdquo Transactionson Intelligent Transportation Systems 2017

[57] S Amini I Gerostathopoulos and C Prehofer ldquoBig Data Analyt-ics Architecture for Real-time Traffic Controlrdquo in 5th InternationalConference on Models and Technologies for Intelligent TransportationSystems (MT-ITS) IEEE 2017 pp 710ndash715

[58] J Yu H Zhu H Han Y J Chen J Yang Y Zhu Z Chen G Xueand M Li ldquoSenspeed Sensing Driving Conditions to Estimate VehicleSpeed in Urban Environmentsrdquo Transactions on Mobile Computingvol 15 no 1 pp 202ndash216 2016

[59] S Nawaz C Efstratiou and C Mascolo ldquoSmart Sensing Systems forthe Daily Driverdquo Pervasive Computing vol 15 no 1 pp 39ndash43 2016

[60] G Han M Guizani Y Bi T H Luan K Ota H Zhou W Guibeneand A Rayes ldquoSoftware-Defined Vehicular Networks ArchitectureAlgorithms and Applications Part 1rdquo Communications Magazinevol 55 no 7 pp 78ndash79 2017

[61] D He S Zeadally B Xu and X Huang ldquoAn Efficient Identity-basedConditional Privacy-preserving Authentication Scheme for VehicularAd Hoc Networksrdquo Transactions on Information Forensics and Secu-rity vol 10 no 12 pp 2681ndash2691 2015

[62] ldquoDeliverable D11 Refined scenarios and requirements consoli-dated use cases and qualitative techno-economic feasibility as-sessmentrdquo httpsmetis-ii5g-pppeuwp-contentuploadsdeliverablesMETIS-II D11 v10pdf 2016 accessed on 18042018

[63] A Osseiran J F Monserrat and P Marsch 5G Mobile and WirelessCommunications Technology Cambridge University Press 2016

[64] S K Datta J Haerri C Bonnet and R F Da Costa ldquoVehicles asConnected Resources Opportunities and Challenges for the FuturerdquoVehicular Technology Magazine vol 12 no 2 pp 26ndash35 2017

[65] V Frascolla F Miatton G K Tran K Takinami A De DomenicoE Calvanese K K Strinati T Haustein K Sakaguchi S Barbarossaet al ldquo5G-MiEdge Design Standardization and Deployment of 5GPhase II Technologiesrdquo in Conference on Standards for Communica-tions amp Networking IEEE 2017 pp 1ndash6

[66] L Li Y Li and R Hou ldquoA Novel Mobile Edge Computing-BasedArchitecture for Future Cellular Vehicular Networksrdquo in WirelessCommunications and Networking Conference (WCNC) IEEE 2017pp 1ndash6

[67] ldquoInformation-Centric Mobile Edge Computing for Connected VehicleEnvironments Challenges and Research Directions author=GreweDennis and Wagner Marco and Arumaithurai Mayutan and PsarasIoannis and Kutscher Dirk booktitle=Proceedings of the Workshopon Mobile Edge Communications pages=7ndash12 year=2017 organiza-tion=ACMrdquo

[68] N H Motlagh M Bagaa and T Taleb ldquoUAV-based IoT platform Acrowd surveillance use caserdquo IEEE Communications Magazine vol 55no 2 pp 128ndash134 2017

[69] M Satyanarayanan ldquoThe emergence of edge computingrdquo Computervol 50 no 1 pp 30ndash39 2017

[70] L Baresi D F Mendonca and M Garriga ldquoEmpowering Low-Latency Applications Through a Serverless Edge Computing Archi-tecturerdquo in European Conference on Service-Oriented and CloudComputing Springer 2017 pp 196ndash210

[71] ldquoETSI executive briefing - mobile edge computing (mec) initiativerdquohttpsportaletsiorgportals0tbpagesmecdocsmec20executive20brief20v12028-09-14pdf accessed on 01022018

[72] M Chen W Saad and C Yin ldquoVirtual Reality Over WirelessNetworks Quality-of-Service Model and Learning-based ResourceManagementrdquo arXiv preprint arXiv170304209 2017

[73] E Bastug M Bennis M Medard and M Debbah ldquoToward Intercon-nected Virtual Reality Opportunities Challenges and Enablersrdquo IEEECommunications Magazine vol 55 no 6 pp 110ndash117 2017

[74] B Cheng G Solmaz F Cirillo E Kovacs K Terasawa and A Ki-tazawa ldquoFogFlow Easy Programming of IoT Services Over Cloud andEdges for Smart Citiesrdquo vol 5 no 2 pp 696ndash707 2018

[75] ldquoCisco Visual Networking Index Forecast and Methodology20162021rdquo Cisco White Paper June 2017 [Online] Availablehttpswwwciscocomcenussolutionscollateralservice-providervisual-networking-index-vnicomplete-white-paper-c11-481360pdf

[76] H Sun Z Zhang R Q Hu and Y Qian ldquoChallenges and En-abling Technologies in 5G Wearable Communicationsrdquo arXiv preprintarXiv170805410 2017

[77] C Perera C H Liu and S Jayawardena ldquoThe Emerging Internetof Things Marketplace from an Industrial Perspective A SurveyrdquoTransactions on Emerging Topics in Computing vol 3 no 4 pp 585ndash598 2015

[78] F J Ferrandez-Pastor J M Garcıa-Chamizo M Nieto-HidalgoJ Mora-Pascual and J Mora-Martınez ldquoDeveloping Ubiquitous Sen-sor Network Platform Using Internet of Things Application in Preci-sion Agriculturerdquo Sensors vol 16 no 7 p 1141 2016

[79] ldquoSmart Farming The sustainable way to foodrdquo Beecham ResearchReport May 2017 accessed on 04042018 [Online] Availablehttpwwwbeechamresearchcom

[80] ldquoBuilding an IoT solution with PeakUp to improve managementof poultry housesrdquo Microsoft Technical Case Studies March2017 [Online] Available httpsmicrosoftgithubiotechcasestudiesiot20170330PeakUphtml

[81] R B Mahale and S Sonavane ldquoSmart Poultry Farm Monitoring UsingIoT and Wireless Sensor Networksrdquo International Journal of AdvancedResearch in Computer Science vol 7 no 3 2016

[82] M Boban K Manolakis M Ibrahim S Bazzi and W Xu ldquoDesignaspects for 5G V2X physical layerrdquo in Conference on Standards forCommunications and Networking (CSCN) IEEE 2016 pp 1ndash7

[83] P J Braun S Pandi R-S Schmoll and F H Fitzek ldquoOn the Studyand Deployment of Mobile Edge Cloud for Tactile Internet usinga 5G Gaming Applicationrdquo in 14th Consumer Communications ampNetworking Conference (CCNC) IEEE 2017 pp 154ndash159

28

[84] S Pandi R S Schmoll P J Braun and F H Fitzek ldquoDemonstrationof Mobile Edge Cloud for Tactile Internet using a 5G Gaming Ap-plicationrdquo in 14th Annual Consumer Communications amp NetworkingConference (CCNC) IEEE 2017 pp 607ndash608

[85] M Satyanarayanan ldquoKeynotes Edge Computing Vision and Chal-lengesrdquo in 2nd International Conference on Collaboration and InternetComputing (CIC) 2016

[86] H Kanzaki K Schubert and N Bambos ldquoVideo Streaming Schemesfor Industrial IoTrdquo in 26th International Conference on ComputerCommunication and Networks (ICCCN) IEEE 2017 pp 1ndash7

[87] K E Harper T de Gooijer J O Schmitt and D Cox ldquoMicrodatabasesfor the industrial internetrdquo arXiv preprint arXiv160104036 2016

[88] G Peralta M Iglesias-Urkia M Barcelo R Gomez A Moran andJ Bilbao ldquoFog Computing Based Efficient IoT Scheme for the Industry40rdquo in International Workshop of Electronics Control MeasurementSignals and their Application to Mechatronics (ECMSM) IEEE 2017pp 1ndash6

[89] J Chakareski ldquoVRAR Immersive Communication Caching EdgeComputing and Transmission Trade-Offsrdquo in Proceedings of theWorkshop on Virtual Reality and Augmented Reality Network ACM2017 pp 36ndash41

[90] A Carvallo and J Cooper The Advanced Smart Grid Edge PowerDriving Sustainability Artech House 2015

[91] M H Y Moghaddam A Leon-Garcia and M Moghaddassian ldquoOnthe Performance of Distributed and Cloud-Based Demand Response inSmart Gridrdquo Transactions on Smart Grid 2017

[92] H Lasi P Fettke H-G Kemper T Feld and M Hoffmann ldquoIndustry40rdquo Business amp Information Systems Engineering vol 6 no 4 pp239ndash242 2014

[93] L Da Xu W He and S Li ldquoInternet of Things in Industries ASurveyrdquo Transactions on industrial informatics vol 10 no 4 pp2233ndash2243 2014

[94] C Perera C H Liu S Jayawardena and M Chen ldquoA Survey onInternet of Things from Industrial Market Perspectiverdquo Access vol 2pp 1660ndash1679 2014

[95] B Kehoe S Patil P Abbeel and K Goldberg ldquoA Survey of Researchon Cloud Robotics and Automationrdquo Transactions on automationscience and engineering vol 12 no 2 pp 398ndash409 2015

[96] J-q Li F R Yu G Deng C Luo Z Ming and Q Yan ldquoIndustrialInternet A Survey on the Enabling Technologies Applications andChallengesrdquo Communications Surveys amp Tutorials 2017

[97] M Albano J B Silva and L Lino Ferreira ldquoThe Industrial Internet ofThingsrdquo 22o Seminario da Rede Tematica de Comunicacoes Moveis2017

[98] R Nelson ldquoSmart Factories Leverage Cloud Edge Computingrdquo EE-Evaluation Engineering vol 56 no 6 pp 14ndash18 2017

[99] W Steiner and S Poledna ldquoFog Computing as Enabler for the Indus-trial Internet of Thingsrdquo e amp i Elektrotechnik und Informationstechnikvol 133 no 7 pp 310ndash314 2016

[100] B Liang Mobile Edge Computing Cambridge University Press 2017[101] D Zhang L T Yang and H Huang ldquoSearching in Internet of Things

Vision and Challengesrdquo in 9th International Symposium on Paralleland Distributed Processing with Applications (ISPA) IEEE 2011pp 201ndash206

[102] P Bellavista and A Zanni ldquoTowards Better Scalability for IoT-cloudInteractions via Combined Exploitation of MQTT and CoAPrdquo in 2ndInternational Forum on Research and Technologies for Society andIndustry Leveraging a better tomorrow (RTSI) IEEE 2016 pp 1ndash6

[103] J Ren H Guo C Xu and Y Zhang ldquoServing at the Edge A ScalableIoT Architecture Based on Transparent Computingrdquo IEEE Networkvol 31 no 5 pp 96ndash105 2017

[104] R Morabito R Petrolo V Loscri and N Mitton ldquoLEGIoT aLightweight Edge Gateway for the Internet of Thingsrdquo Future Gen-eration Computer Systems vol 81 pp 1ndash15 2018

[105] A Ceselli M Premoli and S Secci ldquoMobile Edge Cloud NetworkDesign Optimizationrdquo IEEEACM Transactions on Networking vol 25no 3 pp 1818ndash1831 2017

[106] B Liang Mobile edge computing Cambridge University Press 2017[107] M Peng S Yan K Zhang and C Wang ldquoFog-computing-based radio

access networks issues and challengesrdquo Network vol 30 no 4 pp46ndash53 2016

[108] R Tandon and O Simeone ldquoHarnessing Cloud and Edge SynergiesToward an Information Theory of Fog Radio Access NetworksrdquoCommunications Magazine vol 54 no 8 pp 44ndash50 2016

[109] M Peng and K Zhang ldquoRecent Advances in Fog Radio AccessNetworks Performance Analysis and Radio Resource AllocationrdquoAccess vol 4 pp 5003ndash5009 2016

[110] B P Rimal D P Van and M Maier ldquoMobile-Edge Computingvs Centralized Cloud Computing over a Converged FiWi AccessNetworkrdquo Transactions on Network and Service Management vol 14no 3 pp 498ndash513 2017

[111] M Agiwal A Roy and N Saxena ldquoNext Generation 5G WirelessNetworks A Comprehensive Surveyrdquo Communications Surveys ampTutorials vol 18 no 3 pp 1617ndash1655 2016

[112] S Barbarossa E Ceci M Merluzzi and E Calvanese-StrinatildquoEnabling Effective Mobile Edge Computing Using millimeterwaveLinksrdquo in International Conference on Communications Workshops(ICC Workshops) IEEE 2017 pp 367ndash372

[113] S Barbarossa E Ceci and M Merluzzi ldquoOverbooking Radio andComputation Resources in mmW-Mobile Edge Computing to ReduceVulnerability to Channel Intermittencyrdquo in European Conference onNetworks and Communications (EuCNC) IEEE 2017 pp 1ndash5

[114] A Dongare C Hesling K Bhatia A Balanuta R L PereiraB Iannucci and A Rowe ldquoOpenChirp A Low-Power Wide-AreaNetworking Architecturerdquo in IEEE International Conference on Perva-sive Computing and Communications Workshops (PerCom Workshops)2017 pp 569ndash574

[115] N Ansari and X Sun ldquoMobile Edge Computing Empowers Internetof Thingsrdquo IEICE Transactions on Communications vol 101 no 3pp 604ndash619 2018

[116] I Farris A Orsino L Militano M Nitti G Araniti L Atzoriand A Iera ldquoFederations of Connected Things for Ddelay-sensitiveIoT Services in 5G Environmentsrdquo in International Conference onCommunications (ICC) IEEE 2017 pp 1ndash6

[117] I Farris A Orsino L Militano A Iera and G Araniti ldquoFederated IoTServices Leveraging 5G Technologies at the Edgerdquo Ad Hoc Networksvol 68 pp 58ndash69 2018

[118] A Orsino I Farris L Militano G Araniti S Andreev I GudkovaY Koucheryavy and A Iera ldquoExploiting D2D Communications at theNetwork Edge for Mission-Critical IoT Applicationsrdquo in Proceedingsof 23th European Wireless Conference VDE 2017 pp 1ndash6

[119] ldquoWireless Networks for Mobile Edge Computing Spatial Modeling andLatency Analysis (Extended version) author=Ko Seung-Woo and HanKaifeng and Huang Kaibin journal=arXiv preprint arXiv170901702year=2017rdquo

[120] F Samie V Tsoutsouras L Bauer S Xydis D Soudris and J HenkelldquoComputation Offloading and Resource Allocation for Low-power IoTEdge Devicesrdquo in 3rd World Forum on Internet of Things (WF-IoT)IEEE 2016 pp 7ndash12

[121] S Abdelwahab B Hamdaoui M Guizani and T Znati ldquoReplisomDisciplined Tiny Memory Replication for Massive IoT Devices in LTEEdge Cloudrdquo Internet of Things Journal vol 3 no 3 pp 327ndash3382016

[122] Y Yu X Li and C Qian ldquoSDLB A Scalable and Dynamic SoftwareLoad Balancer for Fog and Mobile Edge Computingrdquo in Proceedingsof the Workshop on Mobile Edge Communications ACM 2017 pp55ndash60

[123] R Vilalta V Lopez A Giorgetti S Peng V Orsini L VelascoR Serral-Gracia D Morris S De Fina F Cugini et al ldquoTelcoFogA Unified Flexible Fog and Cloud Computing Architecture for 5GNetworksrdquo Communications Magazine vol 55 no 8 pp 36ndash43 2017

[124] M Bouet and V Conan ldquoGeo-partitioning of mec resourcesrdquo inProceedings of the Workshop on Mobile Edge Communications ACM2017 pp 43ndash48

[125] H Flores X Su V Kostakos A Y Ding P Nurmi S TarkomaP Hui and Y Li ldquoLarge-scale offloading in the Internet of Thingsrdquoin International Conference on Pervasive Computing and Communica-tions Workshops (PerCom Workshops) IEEE 2017 pp 479ndash484

[126] C Wang C Liang F R Yu Q Chen and L Tang ldquoComputationoffloading and resource allocation in wireless cellular networks withmobile edge computingrdquo IEEE Transactions on Wireless Communica-tions vol 16 no 8 pp 4924ndash4938 2017

[127] X Lyu W Ni H Tian R P Liu X Wang G B Giannakis andA Paulraj ldquoOptimal Schedule of Mobile Edge Computing for Internetof Things Using Partial Informationrdquo Journal on Selected Areas inCommunications vol 35 no 11 pp 2606ndash2615 2017

[128] H Gupta A Vahid Dastjerdi S K Ghosh and R Buyya ldquoiFogSim Atoolkit for modeling and simulation of resource management techniquesin the Internet of Things Edge and Fog computing environmentsrdquoSoftware Practice and Experience vol 47 no 9 pp 1275ndash12962017

[129] K Habak M Ammar K A Harras and E Zegura ldquoFemto CloudsLeveraging Mobile Devices to Provide Cloud Service at the Edgerdquo in

29

8th International Conference on Cloud Computing (CLOUD) IEEE2015 pp 9ndash16

[130] M Chen Y Hao M Qiu J Song D Wu and I Humar ldquoMobility-aware Caching and Computation Offloading in 5G Ultra-dense CellularNetworksrdquo Sensors vol 16 no 7 p 974 2016

[131] X Chen L Jiao W Li and X Fu ldquoEfficient multi-user computationoffloading for mobile-edge cloud computingrdquo IEEEACM Transactionson Networking vol 24 no 5 pp 2795ndash2808 2016

[132] S Sardellitti G Scutari and S Barbarossa ldquoJoint Optimizationof Radio and Computational Resources for Multicell Mobile-edgeComputingrdquo Transactions on Signal and Information Processing overNetworks vol 1 no 2 pp 89ndash103 2015

[133] C Wang F R Yu C Liang Q Chen and L Tang ldquoJoint Compu-tation Offloading and Interference Management in Wireless CellularNetworks With Mobile Edge Computingrdquo Transactions on VehicularTechnology vol 66 no 8 pp 7432ndash7445 2017

[134] Y Sun S Zhou and J Xu ldquoEMM Energy-Aware Mobility Manage-ment for Mobile Edge Computing in Ultra Dense Networksrdquo IEEEJournal on Selected Areas in Communications vol 35 no 11 pp2637ndash2646 2017

[135] J Liu Y Mao J Zhang and K B Letaief ldquoDelay-optimal Com-putation Task Scheduling for Mobile-edge Computing Systemsrdquo inInternational Symposium on Information Theory (ISIT) IEEE 2016pp 1451ndash1455

[136] C You K Huang H Chae and B-H Kim ldquoEnergy-efficient ResourceAllocation for Mobile-edge Computation Offloadingrdquo Transactions onWireless Communications vol 16 no 3 pp 1397ndash1411 2017

[137] P Mach and Z Becvar ldquoCloud-aware Power Control for Cloud-enabledSmall Cellsrdquo in Globecom Workshops IEEE 2014 pp 1038ndash1043

[138] P Mach and Z Becvar ldquoCloud Aware Power Control for Real-timeApplication Offloading in Mobile Edge Computingrdquo Transactions onEmerging Telecommunications Technologies vol 27 no 5 pp 648ndash661 2016

[139] T Taleb and A Ksentini ldquoAn Analytical Model for Follow Me Cloudrdquoin Global Communications Conference (GLOBECOM) IEEE 2013pp 1291ndash1296

[140] D Wu D I Arkhipov E Asmare Z Qin and J A McCann ldquoUbi-Flow Mobility Management in Urban-scale Software Defined IoTrdquo inConference on Computer Communications (INFOCOM) IEEE 2015pp 208ndash216

[141] W Shang A Bannis T Liang Z Wang Y Yu A AfanasyevJ Thompson J Burke B Zhang and L Zhang ldquoNamed DataNetworking of Thingsrdquo in First International Conference on Internet-of-Things Design and Implementation (IoTDI) IEEE 2016 pp 117ndash128

[142] F Giust L Cominardi and C J Bernardos ldquoDistributed mobilityManagement for Future 5G Networks Overview and Analysis ofExisting Approachesrdquo Communications Magazine vol 53 no 1 pp142ndash149 2015

[143] C N Le Tan C Klein and E Elmroth ldquoLocation-aware loadprediction in Edge Data Centersrdquo in Second International Conferenceon Fog and Mobile Edge Computing (FMEC) IEEE 2017 pp 25ndash31

[144] V Vassilakis I P Chochliouros A S Spiliopoulou E SfakianakisM Belesioti N Bompetsis M Wilson C Turyagyenda and A Dard-amanis ldquoSecurity Analysis of Mobile Edge Computing in VirtualizedSmall Cell Networksrdquo in IFIP International Conference on ArtificialIntelligence Applications and Innovations Springer 2016 pp 653ndash665

[145] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems 2016

[146] Q Jing A V Vasilakos J Wan J Lu and D Qiu ldquoSecurity of theInternet of Things Perspectives and Challengesrdquo Wireless Networksvol 20 no 8 pp 2481ndash2501 2014

[147] I Stojmenovic S Wen X Huang and H Luan ldquoAn Overview of FogComputing and its Security Issuesrdquo Concurrency and ComputationPractice and Experience vol 28 no 10 pp 2991ndash3005 2016

[148] J Wan C Zou S Ullah C-F Lai M Zhou and X Wang ldquoCloud-enabled Wireless Body Area Networks for Pervasive HealthcarerdquoNetwork vol 27 no 5 pp 56ndash61 2013

[149] J Wan D Zhang Y Sun K Lin C Zou and H Cai ldquoVCMIA aNovel Architecture for Integrating Vehicular Cyber-physical Systemsand Mobile Cloud Computingrdquo Mobile Networks and Applicationsvol 19 no 2 pp 153ndash160 2014

[150] N Varga L Bokor and E Piri ldquoA Network-assisted Flow MobilityArchitecture for Optimized Mobile Medical Multimedia TransmissionrdquoAnnals of Telecommunications vol 71 no 9-10 pp 489ndash502 2016

[151] M Taylor ldquoThe EU Data Retention Directiverdquo Computer Law ampSecurity Review vol 22 no 4 pp 309ndash312 2006

[152] S Haggard and J R Lindsay ldquoNorth korea and the sony hack export-ing instability through cyberspacerdquo 2015 accessed on 02052018

[153] P German ldquoA New Month a New Data Breachrdquo Network Securityvol 2016 no 3 pp 18ndash20 2016

[154] R Roman J Zhou and J Lopez ldquoOn the Features and Challengesof Security and Privacy in Distributed Internet of Thingsrdquo ComputerNetworks vol 57 no 10 pp 2266ndash2279 2013

[155] F Kemmer C Reich M Knahl and N Clarke ldquoSoftware DefinedPrivacyrdquo in International Conference on Cloud Engineering Workshop(IC2EW) IEEE 2016 pp 25ndash29

[156] S Yi Z Qin and Q Li ldquoSecurity and Privacy Issues of Fog Comput-ing A Surveyrdquo in International Conference on Wireless AlgorithmsSystems and Applications Springer 2015 pp 685ndash695

[157] S F Abedin M G R Alam N H Tran and C S Hong ldquoA FogBased System Model for Cooperative IoT Node Pairing Using Match-ing Theoryrdquo in 17th Asia-Pacific Network Operations and ManagementSymposium (APNOMS) IEEE 2015 pp 309ndash314

[158] P De Hert and V Papakonstantinou ldquoThe Proposed Data ProtectionRegulation replacing Directive 9546EC A Sound System for theProtection of Individualsrdquo Computer Law amp Security Review vol 28no 2 pp 130ndash142 2012

[159] S Ziegler A Skarmeta J Bernal E E Kim and S BianchildquoANASTACIA Advanced Networked Agents for Security and TrustAssessment in CPS IoT Architecturesrdquo in Global Internet of ThingsSummit (GIoTS) IEEE 2017 pp 1ndash6

[160] T D Dang and D Hoang ldquoA Data Protection Model for FogComputingrdquo in Fog and Mobile Edge Computing (FMEC) IEEE2017 pp 32ndash38

[161] R Mijumbi J Serrat J-L Gorricho N Bouten F De Turck andR Boutaba ldquoNetwork Function Virtualization State-of-the-art andResearch Challengesrdquo Communications Surveys amp Tutorials vol 18no 1 pp 236ndash262 2016

[162] L Gupta R Jain and H A Chan ldquoMobileedge computingndashan important ingredient of 5g net-worksrdquo IEEE Software Defined Networks Newsletter2016 [Online] Available httpsdnieeeorgnewslettermarch-2016mobile-edge-computing-an-important-ingredient-of-5g-network

[163] B Yang W K Chai G Pavlou and K V Katsaros ldquoSeamless Supportof Low Latency Mobile Applications with NFV-Enabled Mobile Edge-Cloudrdquo in International Conference on Cloud Networking (Cloudnet)IEEE 2016 pp 136ndash141

[164] B LI Y ZHANG and L XU ldquoAn MEC and NFV Integrated NetworkArchitecturerdquo ZTE COMMUNICATIONS vol 15 no 2 p 1 2017

[165] V Sciancalepore F Giust K Samdanis and Z Yousaf ldquoA Double-tierMEC-NFV Architecture Design and Optimisationrdquo in Conference onStandards for Communications and Networking (CSCN) IEEE 2016pp 1ndash6

[166] G A Carella M Pauls T Magedanz M Cilloni P Bellavista andL Foschini ldquoPrototyping NFV-based Multi-access Edge Computingin 5G Ready Networks with Open Batonrdquo in Conference on NetworkSoftwarization (NetSoft) IEEE 2017 pp 1ndash4

[167] B Blanco J O Fajardo I Giannoulakis E Kafetzakis S PengJ Perez-Romero I Trajkovska P S Khodashenas L GorattiM Paolino et al ldquoTechnology Pillars in the Architecture of Future5G Mobile Networks NFV MEC and SDNrdquo Computer Standards ampInterfaces vol 54 pp 216ndash228 2017

[168] S Peng J O Fajardo P S Khodashenas B Blanco F LiberalC Ruiz C Turyagyenda M Wilson and S Vadgama ldquoQoE-OrientedMobile Edge Service Management Leveraging SDN and NFVrdquo MobileInformation Systems vol 2017 2017

[169] S Ali and M Ghazal ldquoReal-time Heart Attack Mobile DetectionService (RHAMDS) An IoT use case for Software Defined Networksrdquoin 30th Canadian Conference on Electrical and Computer Engineering(CCECE) IEEE 2017 pp 1ndash6

[170] I Farris J Bernabe N Toumi D Garcia-Carrillo T TalebA Skarmeta and B Sahlin ldquoTowards Provisioning of SDNNFV-based Security Enablers for Integrated Protection of IoT Systemsrdquo inConference on Standards for Communications amp Networking (CSCN)IEEE 2017 pp 1ndash6

[171] A Huang N Nikaein T Stenbock A Ksentini and C Bonnet ldquoLowLatency MEC Framework for SDN-based LTELTE-A Networksrdquo inInternational Conference on Communications (ICC) IEEE 2017 pp1ndash6

[172] B Nguyen N Choi M Thottan and J Van der Merwe ldquoSIMECASDN-based IoT Mobile Edge Cloud Architecturerdquo in IFIP Symposium

29

8th International Conference on Cloud Computing (CLOUD) IEEE2015 pp 9ndash16

[130] M Chen Y Hao M Qiu J Song D Wu and I Humar ldquoMobility-aware Caching and Computation Offloading in 5G Ultra-dense CellularNetworksrdquo Sensors vol 16 no 7 p 974 2016

[131] X Chen L Jiao W Li and X Fu ldquoEfficient multi-user computationoffloading for mobile-edge cloud computingrdquo IEEEACM Transactionson Networking vol 24 no 5 pp 2795ndash2808 2016

[132] S Sardellitti G Scutari and S Barbarossa ldquoJoint Optimizationof Radio and Computational Resources for Multicell Mobile-edgeComputingrdquo Transactions on Signal and Information Processing overNetworks vol 1 no 2 pp 89ndash103 2015

[133] C Wang F R Yu C Liang Q Chen and L Tang ldquoJoint Compu-tation Offloading and Interference Management in Wireless CellularNetworks With Mobile Edge Computingrdquo Transactions on VehicularTechnology vol 66 no 8 pp 7432ndash7445 2017

[134] Y Sun S Zhou and J Xu ldquoEMM Energy-Aware Mobility Manage-ment for Mobile Edge Computing in Ultra Dense Networksrdquo IEEEJournal on Selected Areas in Communications vol 35 no 11 pp2637ndash2646 2017

[135] J Liu Y Mao J Zhang and K B Letaief ldquoDelay-optimal Com-putation Task Scheduling for Mobile-edge Computing Systemsrdquo inInternational Symposium on Information Theory (ISIT) IEEE 2016pp 1451ndash1455

[136] C You K Huang H Chae and B-H Kim ldquoEnergy-efficient ResourceAllocation for Mobile-edge Computation Offloadingrdquo Transactions onWireless Communications vol 16 no 3 pp 1397ndash1411 2017

[137] P Mach and Z Becvar ldquoCloud-aware Power Control for Cloud-enabledSmall Cellsrdquo in Globecom Workshops IEEE 2014 pp 1038ndash1043

[138] P Mach and Z Becvar ldquoCloud Aware Power Control for Real-timeApplication Offloading in Mobile Edge Computingrdquo Transactions onEmerging Telecommunications Technologies vol 27 no 5 pp 648ndash661 2016

[139] T Taleb and A Ksentini ldquoAn Analytical Model for Follow Me Cloudrdquoin Global Communications Conference (GLOBECOM) IEEE 2013pp 1291ndash1296

[140] D Wu D I Arkhipov E Asmare Z Qin and J A McCann ldquoUbi-Flow Mobility Management in Urban-scale Software Defined IoTrdquo inConference on Computer Communications (INFOCOM) IEEE 2015pp 208ndash216

[141] W Shang A Bannis T Liang Z Wang Y Yu A AfanasyevJ Thompson J Burke B Zhang and L Zhang ldquoNamed DataNetworking of Thingsrdquo in First International Conference on Internet-of-Things Design and Implementation (IoTDI) IEEE 2016 pp 117ndash128

[142] F Giust L Cominardi and C J Bernardos ldquoDistributed mobilityManagement for Future 5G Networks Overview and Analysis ofExisting Approachesrdquo Communications Magazine vol 53 no 1 pp142ndash149 2015

[143] C N Le Tan C Klein and E Elmroth ldquoLocation-aware loadprediction in Edge Data Centersrdquo in Second International Conferenceon Fog and Mobile Edge Computing (FMEC) IEEE 2017 pp 25ndash31

[144] V Vassilakis I P Chochliouros A S Spiliopoulou E SfakianakisM Belesioti N Bompetsis M Wilson C Turyagyenda and A Dard-amanis ldquoSecurity Analysis of Mobile Edge Computing in VirtualizedSmall Cell Networksrdquo in IFIP International Conference on ArtificialIntelligence Applications and Innovations Springer 2016 pp 653ndash665

[145] R Roman J Lopez and M Mambo ldquoMobile Edge Computing Foget al A Survey and Analysis of Security Threats and ChallengesrdquoFuture Generation Computer Systems 2016

[146] Q Jing A V Vasilakos J Wan J Lu and D Qiu ldquoSecurity of theInternet of Things Perspectives and Challengesrdquo Wireless Networksvol 20 no 8 pp 2481ndash2501 2014

[147] I Stojmenovic S Wen X Huang and H Luan ldquoAn Overview of FogComputing and its Security Issuesrdquo Concurrency and ComputationPractice and Experience vol 28 no 10 pp 2991ndash3005 2016

[148] J Wan C Zou S Ullah C-F Lai M Zhou and X Wang ldquoCloud-enabled Wireless Body Area Networks for Pervasive HealthcarerdquoNetwork vol 27 no 5 pp 56ndash61 2013

[149] J Wan D Zhang Y Sun K Lin C Zou and H Cai ldquoVCMIA aNovel Architecture for Integrating Vehicular Cyber-physical Systemsand Mobile Cloud Computingrdquo Mobile Networks and Applicationsvol 19 no 2 pp 153ndash160 2014

[150] N Varga L Bokor and E Piri ldquoA Network-assisted Flow MobilityArchitecture for Optimized Mobile Medical Multimedia TransmissionrdquoAnnals of Telecommunications vol 71 no 9-10 pp 489ndash502 2016

[151] M Taylor ldquoThe EU Data Retention Directiverdquo Computer Law ampSecurity Review vol 22 no 4 pp 309ndash312 2006

[152] S Haggard and J R Lindsay ldquoNorth korea and the sony hack export-ing instability through cyberspacerdquo 2015 accessed on 02052018

[153] P German ldquoA New Month a New Data Breachrdquo Network Securityvol 2016 no 3 pp 18ndash20 2016

[154] R Roman J Zhou and J Lopez ldquoOn the Features and Challengesof Security and Privacy in Distributed Internet of Thingsrdquo ComputerNetworks vol 57 no 10 pp 2266ndash2279 2013

[155] F Kemmer C Reich M Knahl and N Clarke ldquoSoftware DefinedPrivacyrdquo in International Conference on Cloud Engineering Workshop(IC2EW) IEEE 2016 pp 25ndash29

[156] S Yi Z Qin and Q Li ldquoSecurity and Privacy Issues of Fog Comput-ing A Surveyrdquo in International Conference on Wireless AlgorithmsSystems and Applications Springer 2015 pp 685ndash695

[157] S F Abedin M G R Alam N H Tran and C S Hong ldquoA FogBased System Model for Cooperative IoT Node Pairing Using Match-ing Theoryrdquo in 17th Asia-Pacific Network Operations and ManagementSymposium (APNOMS) IEEE 2015 pp 309ndash314

[158] P De Hert and V Papakonstantinou ldquoThe Proposed Data ProtectionRegulation replacing Directive 9546EC A Sound System for theProtection of Individualsrdquo Computer Law amp Security Review vol 28no 2 pp 130ndash142 2012

[159] S Ziegler A Skarmeta J Bernal E E Kim and S BianchildquoANASTACIA Advanced Networked Agents for Security and TrustAssessment in CPS IoT Architecturesrdquo in Global Internet of ThingsSummit (GIoTS) IEEE 2017 pp 1ndash6

[160] T D Dang and D Hoang ldquoA Data Protection Model for FogComputingrdquo in Fog and Mobile Edge Computing (FMEC) IEEE2017 pp 32ndash38

[161] R Mijumbi J Serrat J-L Gorricho N Bouten F De Turck andR Boutaba ldquoNetwork Function Virtualization State-of-the-art andResearch Challengesrdquo Communications Surveys amp Tutorials vol 18no 1 pp 236ndash262 2016

[162] L Gupta R Jain and H A Chan ldquoMobileedge computingndashan important ingredient of 5g net-worksrdquo IEEE Software Defined Networks Newsletter2016 [Online] Available httpsdnieeeorgnewslettermarch-2016mobile-edge-computing-an-important-ingredient-of-5g-network

[163] B Yang W K Chai G Pavlou and K V Katsaros ldquoSeamless Supportof Low Latency Mobile Applications with NFV-Enabled Mobile Edge-Cloudrdquo in International Conference on Cloud Networking (Cloudnet)IEEE 2016 pp 136ndash141

[164] B LI Y ZHANG and L XU ldquoAn MEC and NFV Integrated NetworkArchitecturerdquo ZTE COMMUNICATIONS vol 15 no 2 p 1 2017

[165] V Sciancalepore F Giust K Samdanis and Z Yousaf ldquoA Double-tierMEC-NFV Architecture Design and Optimisationrdquo in Conference onStandards for Communications and Networking (CSCN) IEEE 2016pp 1ndash6

[166] G A Carella M Pauls T Magedanz M Cilloni P Bellavista andL Foschini ldquoPrototyping NFV-based Multi-access Edge Computingin 5G Ready Networks with Open Batonrdquo in Conference on NetworkSoftwarization (NetSoft) IEEE 2017 pp 1ndash4

[167] B Blanco J O Fajardo I Giannoulakis E Kafetzakis S PengJ Perez-Romero I Trajkovska P S Khodashenas L GorattiM Paolino et al ldquoTechnology Pillars in the Architecture of Future5G Mobile Networks NFV MEC and SDNrdquo Computer Standards ampInterfaces vol 54 pp 216ndash228 2017

[168] S Peng J O Fajardo P S Khodashenas B Blanco F LiberalC Ruiz C Turyagyenda M Wilson and S Vadgama ldquoQoE-OrientedMobile Edge Service Management Leveraging SDN and NFVrdquo MobileInformation Systems vol 2017 2017

[169] S Ali and M Ghazal ldquoReal-time Heart Attack Mobile DetectionService (RHAMDS) An IoT use case for Software Defined Networksrdquoin 30th Canadian Conference on Electrical and Computer Engineering(CCECE) IEEE 2017 pp 1ndash6

[170] I Farris J Bernabe N Toumi D Garcia-Carrillo T TalebA Skarmeta and B Sahlin ldquoTowards Provisioning of SDNNFV-based Security Enablers for Integrated Protection of IoT Systemsrdquo inConference on Standards for Communications amp Networking (CSCN)IEEE 2017 pp 1ndash6

[171] A Huang N Nikaein T Stenbock A Ksentini and C Bonnet ldquoLowLatency MEC Framework for SDN-based LTELTE-A Networksrdquo inInternational Conference on Communications (ICC) IEEE 2017 pp1ndash6

[172] B Nguyen N Choi M Thottan and J Van der Merwe ldquoSIMECASDN-based IoT Mobile Edge Cloud Architecturerdquo in IFIP Symposium

30

on Integrated Network and Service Management (IM) IEEE 2017pp 503ndash509

[173] M S Hossain C Xu Y Li A-S K Pathan J Bilbao W Zeng andA El Saddik ldquoImpact of Next-Generation Mobile Technologies onIoT-Cloud Convergencerdquo Communications Magazine vol 55 no 1pp 18ndash19 2017

[174] J Liu J Wan D Jia B Zeng D Li C-H Hsu and H Chen ldquoHigh-Efficiency Urban Traffic Management in Context-Aware Computingand 5G Communicationrdquo Communications Magazine vol 55 no 1pp 34ndash40 2017

[175] K Phemius J Seddar M Bouet H Khalife and V Conan ldquoBringingSDN to the Edge of Tactical Networksrdquo in Military CommunicationsConference (MILCOM) IEEE 2016 pp 1047ndash1052

[176] C Aggarwal and K Srivastava ldquoSecuring IoT devices using SDN andedge computingrdquo in 2nd International Conference on Next GenerationComputing Technologies (NGCT) IEEE 2016 pp 877ndash882

[177] A V Vasilakos Z Li G Simon and W You ldquoInformation centricnetwork Research challenges and opportunitiesrdquo Journal of Networkand Computer Applications vol 52 pp 1ndash10 2015

[178] G Piro L A Grieco G Boggia and P Chatzimisios ldquoInformation-centric Networking and Multimedia Services Present and Future Chal-lengesrdquo Transactions on Emerging Telecommunications Technologiesvol 25 no 4 pp 392ndash406 2014

[179] E Ahmed M Imran M Guizani A Rayes J Lloret G Han andW Guibene ldquoEnabling Mobile and Wireless Technologies for SmartCitiesrdquo IEEE Communications Magazine vol 55 no 1 pp 74ndash752017

[180] M Maier M Chowdhury B P Rimal and D P Van ldquoThe TactileInternet Vision Recent Progress and Open Challengesrdquo Communica-tions Magazine vol 54 no 5 pp 138ndash145 2016

[181] R Ravindran A Chakraborti S O Amin A Azgin and G WangldquoRealizing ICN in 3GPPrsquos 5G NextGen Core Architecturerdquo arXivpreprint arXiv171102232 2017

[182] ldquoUnderstanding Information-Centric Networking and Mo-bile Edge Computingrdquo 5G Americas Decem-ber 2016 accessed on 12012018 [Online] Avail-able httpwww5gamericasorgfiles121481753330UnderstandingInformation Centric Networking and Mobile Edge Computingpdf

[183] Y Zhou and F R Yu and J Chen and Y Kuo ldquoVideo TranscodingCaching and Multicast for Heterogeneous Networks over WirelessNetwork Virtualizationrdquo Communications Letters vol 22 no 1 pp141ndash144 2018

[184] Y Zhou F R Yu J Chen and Y Kuo ldquoResource Allocation forInformation Centric Virtualized Heterogeneous Networks with In-Network Caching and Mobile Edge Computingrdquo Transactions onVehicular Technology vol 66 no 12 pp 11 339ndash11 351 2017

[185] R Huo F R Yu T Huang R Xie J Liu V C Leung and Y LiuldquoSoftware Defined Networking Caching and Computing for GreenWireless Networksrdquo Communications Magazine vol 54 no 11 pp185ndash193 2016

[186] C Ge N Wang S Skillman G Foster and Y Cao ldquoQoE-DrivenDASH Video Caching and Adaptation at 5G Mobile Edgerdquo in Pro-ceedings of 3rd ACM Conference on Information-Centric NetworkingACM 2016 pp 237ndash242

[187] K Samdanis X Costa-Perez and V Sciancalepore ldquoFrom NetworkSharing to Multi-tenancy The 5G Network Slice Brokerrdquo Communi-cations Magazine vol 54 no 7 pp 32ndash39 2016

[188] N Alliance ldquoDescription of Network Slicing Conceptrdquo NGMN 5G Pvol 1 2016

[189] N Nikaein E Schiller R Favraud K Katsalis D StavropoulosI Alyafawi Z Zhao T Braun and T Korakis ldquoNetwork StoreExploring Slicing in Future 5G Networksrdquo in Proceedings of the10th International Workshop on Mobility in the Evolving InternetArchitecture ACM 2015 pp 8ndash13

[190] ldquoNetwork Slicing for 5G Networks amp Servicesrdquo 5G Americas WhitePaper Network Slicing for 5G and Beyond November 2015 accessedon 03012018 [Online] Available httpwww5gamericasorgfiles3214797501045G Americas Network Slicing 1121 Finalpdf

[191] H Zhang N Liu X Chu K Long A-H Aghvami and V C LeungldquoNetwork Slicing Based 5G and Future Mobile Networks MobilityResource Management and Challengesrdquo Communications Magazinevol 55 no 8 pp 138ndash145 2017

[192] K Katsalis N Nikaein E Schiller A Ksentini and T Braun ldquoNet-work Slices toward 5G Communications Slicing the LTE NetworkrdquoCommunications Magazine vol 55 no 8 pp 146ndash154 2017

[193] R Munoz L Nadal R Casellas M S Moreolo R VilaltaJ M Fabrega R Martınez A Mayoral and F J Vılchez ldquoThe

ADRENALINE testbed An SDNNFV packetoptical transport net-work and edgecore cloud platform for end-to-end 5G and IoT ser-vicesrdquo in European Conference on Networks and Communications(EuCNC) IEEE 2017 pp 1ndash5

[194] F van Lingen M Yannuzzi A Jain R Irons-Mclean O LluchD Carrera J L Perez A Gutierrez D Montero J Marti et alldquoThe Unavoidable Convergence of NFV 5G and Fog A Model-DrivenApproach to Bridge Cloud and Edgerdquo Communications Magazinevol 55 no 8 pp 28ndash35 2017

[195] R Vilalta A Mayoral R Casellas R Martınez and R MunozldquoSDNNFV Orchestration of Multi-technology and Multi-domain Net-works in CloudFog Architectures for 5G Servicesrdquo in 21st Opto-Electronics and Communications Conference (OECC) held jointly with2016 International Conference on Photonics in Switching (PS) IEEE2016 pp 1ndash3

[196] R Ravindran A Chakraborti S O Amin A Azgin and G Wangldquo5G-ICN Delivering ICN Services over 5G Using Network SlicingrdquoCommunications Magazine vol 55 no 5 pp 101ndash107 2017

[197] ldquoSESAME Projectrdquo H2020 EU project accessed on 25032018[Online] Available httpwwwsesame-h2020-5g-pppeuHomeaspx

[198] ldquoANASTACIA Projectrdquo H2020 EU project accessed on 11022018[Online] Available httpwwwanastacia-h2020eu

[199] ldquo5G-MiEdge project Millimeter-wave Edge Cloud as an Enablerfor 5G Ecosystemrdquo H2020 EUampJapan Project 2017 accessed on15022018 [Online] Available https5g-miedgeeu

[200] ldquo5GPagodardquo EU Japan collaboration project accessed on 19022018[Online] Available https5g-pagodaaaltofi

[201] ldquoInter-IoT Projectrdquo H2020 EU project accessed on 15022018[Online] Available httpwwwinter-iot-projecteu

[202] ldquo5G MoNArch Projectrdquo H2020 EU project accessed on 17022018[Online] Available https5g-monarcheu

[203] ldquo5G ESSENCE Projectrdquo H2020 EU project accessed on 15042018[Online] Available https5g-pppeu5g-essence

[204] ldquoMATILDA Projectrdquo H2020 EU project accessed on 15042018[Online] Available https5g-pppeumatilda

[205] ldquo5GCity Projectrdquo H2020 EU project accessed on 22022018[Online] Available httpwww5gcityeu

[206] ldquoMONICA Projectrdquo H2020 EU project accessed on 11032018[Online] Available httpwwwmonica-projecteu

[207] ldquoAUTOPILOT Projectrdquo H2020 EU project accessed on 15022018[Online] Available httpautopilot-projecteu

[208] ldquo5G-CORAL Projectrdquo H2020 EU project accessed on 15032018[Online] Available http5g-coraleu

[209] ldquoETSI and VRARA cooperate on Virtual and Augmented RealityrdquoETSI news event accessed on 04052018 [Online] Availablehttpwwwetsiorgnews-events

[210] J Liu T Zhao S Zhou Y Cheng and Z Niu ldquoCONCERT acloud-based architecture for next-generation cellular systemsrdquo WirelessCommunications vol 21 no 6 pp 14ndash22 2014

[211] A Mestres A Rodriguez-Natal J Carner P Barlet-Ros E AlarconM Sole V Muntes-Mulero D Meyer S Barkai M J Hibbett et alldquoKnowledge-defined Networkingrdquo SIGCOMM Computer Communica-tion Review vol 47 no 3 pp 2ndash10 2017

[212] A Crutcher C Koch K Coleman J Patman F Esposito andP Calyam ldquoHyperprofile-based Computation Offloading for MobileEdge Networksrdquo arXiv preprint arXiv170709422 2017

[213] E Ahmed A Ahmed I Yaqoob J Shuja A Gani M Imran andM Shoaib ldquoBringing Computation Closer toward the User NetworkIs Edge Computing the Solutionrdquo IEEE Communications Magazinevol 55 no 11 pp 138ndash144 2017

[214] L T Sorensen S Khajuria and K E Skouby ldquo5G Visions ofUser Privacyrdquo in 81st Vehicular Technology Conference (VTC Spring)IEEE 2015 pp 1ndash4

[215] A Cavoukian and M Chibba ldquoA Regulartorrsquos Perspective Leadingthe way with Privacy by Designrdquo Cyber security in future Internetsecurity and privacy by design OUTLOOK Visions and research forthe wireless world no 11 2014

[216] D Pitt ldquoTrust in the Cloud The Role of SDNrdquo Network Security vol2013 no 3 pp 5ndash6 2013

[217] J Aikat A Akella J S Chase A Juels M K Reiter T RistenpartV Sekar and M Swift ldquoRethinking Security in the Era of CloudComputingrdquo Security amp Privacy vol 15 no 3 pp 60ndash69 2017

[218] H Li G Shou Y Hu and Z Guo ldquoMobile Edge ComputingProgress and Challengesrdquo in International Conference on Mobile CloudComputing Services and Engineering (MobileCloud) IEEE 2016pp 83ndash84

31

Pawani Porambage Pawani Porambage is a PhDstudent the Centre for Wireless CommunicationsUniversity of Oulu Finland She obtained her Bach-elor Degree in Electronics and TelecommunicationEngineering in 2010 from University of MoratuwaSri Lanka and her Masters Degree in UbiquitousNetworking and Computer Networking in 2012 fromUniversity of Nice Sophia-Anipolis France Hermain research interests include lightweight securityprotocols security and privacy on IoT and MECand Wireless Sensor Networks

Jude Okwuibe Jude Okwuibe received his BScin Telecommunications and Wireless Technologiesfrom the American University of Nigeria Yola in2011 In 2015 Okwuibe received a master degreein Wireless Communications Engineering from theUniversity of Oulu Finland Jude is currently doinga doctoral programme in Communications Engi-neering at the University of Oulu Graduate School(UniOGS) Finland His research interests are 5Gand future networks IoT SDN Network securityand biometric verifications

Madhusanka Liyanage received his BSc degree(First Class Honours) in electronics and telecom-munication engineering from the University ofMoratuwa Moratuwa Sri Lanka in 2009 theMEng degree from the Asian Institute of Technol-ogy Bangkok Thailand in 2011 the MSc degreefrom the University of Nice Sophia Antipolis NiceFrance in 2011 and the PhD degree in commu-nication engineering from the University of OuluOulu Finland in 2016 From 2011 to 2012 heworked a Research Scientist at the I3S Laboratory

and Inria Shopia Antipolis France He is currently a Post-Doctoral Researcherand a Project Manager at the Center for Wireless Communications Universityof Oulu He has been a Visiting Research Fellow at the Department ofComputer Science University of Oxford Data61 CSIRO Sydney Australiathe Infolabs21 Lancaster University UK and Computer Science and Engi-neering The University of New South Wales during 2015-2018

He has co-authored over 40 publications including two edited books withWiley and one patent He served as a Technical program Committee Membersat EAI M3Apps 2016 5GU 2017 EUCNC 2017 EUCNC 2018 MASS2018 5G-WF 2018 MCWN 2018 conferences and Technical program co-chair in SecureEdge workshop at IEEE CIT2017 MEC-IoT Workshop at5GWF 2018 and BlockchainIoT workshop at Globecom 2018 conferencesHe has also served as the session chair in a number of other conferencesincluding IEEE WCNC 2013 CROWNCOM 2014 5GU 2014 IEEE CIT2017 IEEE PIMRC 2017 Moreover He has received two best Paper Awardsin the areas of SDMN security (at NGMAST 2015) and 5G Security (atIEEE CSCN 2017) Additionally he has been awarded two research grants(IRC Postdoctoral Grant and Marie-Curie Fellowship) and 21 other prestigiousawardsscholarships during his research career

Dr Liyanage has worked for more than twelve EU international andnational projects in ICT domain He held responsibilities as a leader ofwork packages in several national and EU projects Currently he is theFinnish national coordinator for EU COST Action CA15127 on resilientcommunication services In addition he iswas serving as a managementcommittee member for four other EU COST action projects namely EU COSTAction IC1301 IC1303 CA15107 and CA16226 Liyanage has over threeyears experience in research project management research group leadershipresearch project proposal preparation project progress documentation andgraduate student co-supervisionmentoring skills In 2015 2016 and 2017 hewon the Best Researcher Award at the Centre for Wireless CommunicationsUniversity of Oulu for his excellent contribution in project management anddissemination activities Additionally two of the research projects (MEVICOand SIGMONA projects) received the CELTIC Excellence Award in 2013 and2017 respectively

Dr Liyanagersquos research interests are SDN IoT Blockchain MEC mobileand virtual network security Contact him at madhusankaliyanageoulufi

Mika Ylianttila Prof Mika Ylianttila is a full-timeprofessor at the Centre for Wireless Communications(CWC) at the Faculty of Information Technologyand Electrical Engineering (ITEE) University ofOulu Finland Previously he was the director ofthe Center for Internet Excellence (20122015) andassociate director of the MediaTeam research group(20092011) and professor (pro tem) in Informationnetworks (20052010) He is also adjunct professor inComputer Science and Engineering (since 2007) Hereceived his doctoral degree on Communications En-

gineering at the University of Oulu in 2005 He has coauthored more than 100international peer-reviewed articles on broadband communications networksand systems including aspects on network security mobility managementdistributed systems and novel applications Research Interests include also 5Gapplications and services SDN and edge computing He is a Senior Memberof IEEE and Editor in Wireless Networks journal

Tarik Taleb Prof Tarik Taleb is an IEEE Commu-nications Society (ComSoc) Distinguished Lecturerand a senior member of IEEE He is currently Pro-fessor at the School of Electrical Engineering AaltoUniversity Finland Prior to his current academicposition he was working as Senior Researcher and3GPP Standards Expert at NEC Europe Ltd Hei-delberg Germany He was then leading the NECEurope Labs Team working on RampD projects oncarrier cloud platforms an important vision of 5Gsystems Before joining NEC and till Mar 2009

he worked as assistant professor at the Graduate School of InformationSciences Tohoku University Japan in a lab fully funded by KDDI From Oct2005 till Mar 2006 he worked as research fellow at the Intelligent CosmosResearch Institute Sendai Japan He received his B E degree in InformationEngineering with distinction MSc and PhD degrees in Information Sciencesfrom Tohoku Univ in 2001 2003 and 2005 respectively

Prof Talebs research interests lie in the field of architectural enhancementsto mobile core networks (particularly 3GPPs) mobile cloud networkingnetwork function virtualization software defined networking mobile multime-dia streaming inter-vehicular communications and social media networkingProf Taleb has been also directly engaged in the development and standardiza-tion of the Evolved Packet System as a member of 3GPPs System Architectureworking group Prof Taleb is a member of the IEEE Communications SocietyStandardization Program Development Board As an attempt to bridge thegap between academia and industry Prof Taleb founded the IEEE Workshopon Telecommunications Standards from Research to Standards a successfulevent that got awarded best workshop award by IEEE Communication Society(ComSoC) Based on the success of this workshop Prof Taleb has alsofounded and has been the steering committee chair of the IEEE Conf onStandards for Communications and Networking

Prof Taleb is the general chair of the 2019 edition of the IEEE WirelessCommunications and Networking Conference (WCNC19) to be held in Mar-rakech Morocco He iswas on the editorial board of the IEEE Transactions onWireless Communications IEEE Wireless Communications Magazine IEEEJournal on Internet of Things IEEE Transactions on Vehicular TechnologyIEEE Communications Surveys amp Tutorials and a number of Wiley journalsTill Dec 2016 he served as chair of the Wireless Communications TechnicalCommittee the largest in IEEE ComSoC He also served as Vice Chair of theSatellite and Space Communications Technical Committee of IEEE ComSoc(2006 - 2010) He has been on the technical program committee of differentIEEE conferences including Globecom ICC and WCNC and chaired someof their symposia

Prof Taleb is the (co)recipient of the 2017 IEEE Communications SocietyFred W Ellersick Prize (May 2017) the 2009 IEEE ComSoc Asia-PacificBest Young Researcher award (Jun 2009) the 2008 TELECOM SystemTechnology Award from the Telecommunications Advancement Foundation(Mar 2008) the 2007 Funai Foundation Science Promotion Award (Apr2007) the 2006 IEEE Computer Society Japan Chapter Young Author Award(Dec 2006) the Niwa Yasujirou Memorial Award (Feb 2005) and theYoung Researcherrsquos Encouragement Award from the Japan chapter of theIEEE Vehicular Technology Society (VTS) (Oct 2003) Some of Prof Talebsresearch work have been also awarded best paper awards at prestigiousconferences

• I Introduction
• • I-A Role of MEC for IoT
  • I-B Paper motivation
  • I-C Paper organization
  • • II IoT and MEC application scenarios
    • • II-A Smart home and Smart city
      • II-B Healthcare
      • II-C Autonomous VehiclesIoT Automotive
      • II-D Gaming AR and VR
      • II-E Retail
      • II-F Wearable IoT (WIoT)
      • II-G IoT in Mechanized Agriculture
      • II-H Smart Energy
      • II-I Industrial Internet
      • • III Technical Aspects of MEC Enabled IoT
        • • III-A Scalability
          • • III-A1 Requirements
            • III-A2 Related work
            • • III-B Communication
              • • III-B1 Requirements
                • III-B2 Related work
                • • III-C Computation Offloading and Resource Allocation
                  • • III-C1 Requirements
                    • III-C2 Related work
                    • • III-D Mobility Management
                      • • III-D1 Requirements
                        • III-D2 Related Work
                        • • III-E Security
                          • • III-E1 Requirements
                            • III-E2 Related Work
                            • • III-F Privacy
                              • • III-F1 Issues and challenges
                                • III-F2 Related work
                                • • III-G Trust management
                                  • • III-G1 Requirements
                                    • III-G2 Related work
                                    • • IV Integration Technologies
                                      • • IV-A Network Function Virtualization
                                        • IV-B Software Defined Networking
                                        • IV-C Information Centric Networking
                                        • IV-D Network Slicing
                                        • • V Projects
                                          • • V-1 SESAME Small cEllS coordinAtion for Multi-tenancy and Edge services (June 2015 - Dec 2017)
                                            • • V-2 ANASTACIA Advanced Networked Agents for Security and Trust Assessment in CPS IOT Architectures (Jan 2017 - Dec 2019)
                                              • V-3 5G-MiEdge Millimeter-wave Edge Cloud as an Enabler for 5G Ecosystem (July 2016 - June 2019)
                                              • V-4 5GPagoda
                                              • V-5 Inter-IoT (Jan 2016 - Dec 2018)
                                              • V-6 5G-MoNArch 5G Mobile Network Architecture for diverse services use cases and applications in 5G and beyond (July 2017 - June 2019)
                                              • V-7 5G-ESSENSE Embedded Network Services for 5G Experiences (June 2017 - June 2019)
                                              • V-8 MATILDA (June 2017 - June 2019)
                                              • V-9 5GCITY (June 2017 - June 2019)
                                              • V-10 MONICA Management Of Networked IoT Wearables acircbullfi Very Large Scale Demonstration of Cultural and Societal Applications (Jan 2017 - Dec 2019)
                                              • V-11 AUTOPILOT AUTOmated driving Progressed by Internet Of Things (Jan 2017 - Dec 2019)
                                              • V-12 5G-CORAL A 5G Convergent Virtualised Radio Access Network Living at the Edge (Sep 2017 - Aug 2019)
                                              • • VI Lessons Learned and Future Research Directions
                                                • • VI-A Applications
                                                  • • VI-A1 Lessons learned
                                                    • VI-A2 Future research directions
                                                    • • VI-B Scalability
                                                      • • VI-B1 Lessons learned
                                                        • VI-B2 Future research directions
                                                        • • VI-C Communication
                                                          • • VI-C1 Lessons learned
                                                            • VI-C2 Future research directions
                                                            • • VI-D Computation Offloading and Resource Allocation
                                                              • • VI-D1 Lessons learned
                                                                • VI-D2 Future research directions
                                                                • • VI-E Mobility Management
                                                                  • • VI-E1 Lessons learned
                                                                    • VI-E2 Future research directions
                                                                    • • VI-F Security
                                                                      • • VI-F1 Lessons learned
                                                                        • VI-F2 Future research directions
                                                                        • • VI-G Privacy
                                                                          • • VI-G1 Lessons learned
                                                                            • VI-G2 Future research directions
                                                                            • • VI-H Trust Management
                                                                              • • VI-H1 Lessons learned
                                                                                • VI-H2 Future research directions
                                                                                • • VI-I Standardization
                                                                                  • • VI-I1 Future research directions
                                                                                    • • VII Conclusions
                                                                                      • References
                                                                                      • Biographies
                                                                                      • • Pawani Porambage
                                                                                        • Jude Okwuibe
                                                                                        • Madhusanka Liyanage
                                                                                        • Mika Ylianttila
                                                                                        • Tarik Taleb