On the Integration of Cloud Computingand Internet of ThingsAlessio Botta, Walter de Donato, Valerio Persico, Antonio Pescap eUniversity of Napoli Federico II{a.botta, walter.dedonato, valerio.persico, pescape}@unina.itAbstractCloud computing and Internet of Things (IoT), twovery different technologies, are both already part of our life. Theirmassive adoption and use is expected to increase further, makingthemimportant components of the Future Internet. Anovelparadigm where Cloud and IoT are merged together is foreseenasdisruptiveandanenablerofalargenumberof applicationscenarios. In this paper we focus our attention on the integrationof Cloud and IoT, which we call the CloudIoTparadigm. Manyworks inliterature have surveyedCloudandIoTseparately:their main properties, features, underlying technologies, and openissues. However, to the best of our knowledge, these workslackadetailedanalysis of theCloudIoTparadigm. Tobridgethis gap, in this paper we reviewthe literature about theintegration of Cloud and IoT. We start analyzing and discussingthe need for integrating them, the challenges deriving from suchintegration, and how these issues have been tackled in literature.We then describe application scenarios that have been presentedinliterature, aswell asplatformsbothcommercial andopensource andprojects implementing the CloudIoTparadigm.Finally, we identify open issues, main challenges and futuredirections in this promising eld.I. INTRODUCTION AND MOTIVATIONThe Internet of Things (IoT) paradigmis basedonin-telligent andself conguringnodes (things) interconnectedinadynamicandglobalnetworkinfrastructure. Itrepresentsoneofthemost disruptivetechnologies, enablingubiquitousandpervasivecomputingscenarios. IoTisgenerallycharac-terizedbyreal worldandsmall thingswithlimitedstorageandprocessingcapacity, andconsequential issues regardingreliability, performance, security, andprivacy. Onthe otherhand, Cloud computing has virtually unlimited capabilities interms of storage and processing power, is a much more maturetechnology, andhasmost oftheIoTissuesat least partiallysolved. Thus, a novel IT paradigm in which Cloud and IoT aretwo complementary technologies merged together is expectedto disrupt both current and future Internet [59], [23]. We callthis new paradigm CloudIoT. This paper reviews the literatureabout the integrationof CloudandIoTa reallypromisingtopicfor bothresearchandindustry. Wehavereviewedtherich and articulate state of the art in this eld by consideringalargenumber of papers proposinganintegratedusageofCloud and IoT, and published between 2008 and 2013 indifferent, selectedvenues. As showninFig. 1, bothtopicsgained popularity in the last fewyears (Fig. 1a), and thenumber of papers dealing with Cloud and IoT separately showsan increasing trend since 2008 (Fig. 1b)1. On the other hand,1DatahavebeenobtainedfromGoogleTrends(https://www.google.com/trends/) and Scholar (scholar.google.com/) web facilities.amorerecent andrapidlyincreasingtrenddealswithCloudandIoTtogether. Followingtheindicationsreportedin[40],we adopt the research methodology schematically depicted inFig. 2. We rst provide a temporal characterizationof theliterature aiming at showing in a qualitative way the temporalbehavioroftheresearchandthecommoninterest about theCloudIoT paradigm. Second, we provide a detailed discussionontheCloudIoTparadigm,highlightingthecomplementarityandtheneedfor their integration. Third, wedetail thenewapplication scenarios stemming fromthe adoption of theCloudIoTparadigm. Fourth, jointlyanalyzingtheCloudIoTparadigmand the application scenarios, we derive the hottopics and related issues for research. Fifth, we describethemainplatforms(bothcommercial andopensource) andresearchprojectsintheeldofCloudIoT. Finally, thankstothe previous seven steps, we derive the open issues and futuredirections in the eld of CloudIoT.(a) Interest from Google research trends. (b) Interest by content and title.Fig. 1: Research and interest trends about Cloud and IoT.II. CLOUD AND IOT: THE NEED FOR THEIRINTEGRATIONThe two worlds of Cloud and IoT have seen an independentevolution.However,severalmutualadvantagesderivingfromtheir integration have been identied in literature and are fore-seen in the future. On the one hand, IoT can benet from thevirtually unlimited capabilities and resources of Cloud to com-pensate its technological constraints (e.g., storage, processing,energy). Specically, the Cloud can offer an effective solutionto implement IoT service management and composition as wellas applications that exploit thethings or thedataproducedby them. On the other hand, the Cloud can benet fromIoTbyextendingitsscopetodeal withreal worldthingsinamoredistributedanddynamicmanner, andfor deliveringnewservices inalargenumber of real lifescenarios. Thecomplementary characteristics of Cloud and IoT arising fromthe different proposals in literature and inspiring the CloudIoTFig. 2: The research methodology adopted in this work.paradigmarereportedinTab. I. Essentially, theCloudactsas intermediate layer between the things and the applications,where it hides all the complexity and the functionalitiesnecessary to implement the latter. This framework will impactfutureapplicationdevelopment,whereinformationgathering,processing, andtransmissionwill producenewchallengestobeaddressed, alsoinamulti-cloudenvironment [30]. Inthefollowing, we summarize the issues solved and the advantagesobtained when adopting the CloudIoT paradigm.Storage resources. IoT involves by denition a large amountof information sources (i.e., the things), which produce a hugeamount of non-structured or semi-structured data [30] havingthe three characteristics typical of Big Data [60]: volume (i.e.,datasize), variety(i.e., datatypes), andvelocity(i.e., datageneration frequency). Hence it implies collecting, accessing,processing, visualizing, archiving, sharing, and searching largeamountsofdata[50]. Offeringvirtuallyunlimited, low-cost,and on-demand storage capacity, Cloud is the most convenientand cost effective solution to deal with data produced byIoT[50]. This integrationrealizes a newconvergence sce-nario[47], wherenewopportunities arisefor dataaggrega-tion [32], integration [56], and sharing with third parties [56].OnceintotheCloud, datacanbetreatedinahomogeneousmanner through standard APIs [32], can be protected byapplying top-level security [27], and directly accessed andvisualized from any place [50].Computational resources. IoT devices have limited process-ingresourcesthatdonotallowon-sitedataprocessing. Datacollected is usually transmitted to more powerful nodes whereaggregation and processing is possible, but scalability is chal-lenging to achieve without a proper infrastructure. The unlim-ited processing capabilities of Cloud and its on-demand modelallow IoT processing needs to be properly satised and enableanalyses of unprecedented complexity [27], [47]. Data-drivendecision making and prediction algorithms would be possibleat low cost and would provide increasing revenues and reducedrisks[56]. Otherperspectiveswouldbetoperformreal-timeprocessing (on-the-y) [50], [27], to implement scalable, real-time, collaborative, sensor-centric applications [32], to managecomplexevents [50], andtoimplement taskofoadingforenergy saving [53].Communicationresources. OneoftherequirementsofIoTTABLE I: Complementarity and Integration of Cloud and IoT.IoT Cloudpervasive ubiquitous (resources usable(things placed everywhere) from everywhere)real world things virtual resourceslimited virtually unlimitedcomputational capabilities computational capabilitieslimited storage or virtually unlimitedno storage capabilities storage capabilitiesInternet as a point of convergence Internet for service deliverybig data source means to manage big datais to make IP-enabled devices communicate through dedicatedhardware, andthe support for suchcommunicationcanbeveryexpensive. Cloudoffersaneffectiveandcheapsolutiontoconnect, track, andmanageanythingfromanywhereatany time using customized portals and built-in apps [50].Thankstotheavailabilityofhighspeednetworks, itenablesthemonitoringandcontrolofremotethings[50], [32], [47],their coordination [32], [47], [52], their communications [32],and the real-time access to the produced data [50].Newcapabilities. IoTis characterizedbyaveryhighhet-erogeneity of devices, technologies, and protocols. Therefore,scalability, interoperability, reliability, efciency, availability,andsecuritycanbeverydifcult toobtain. Theintegrationwith the Cloud solves most of these problems [32], [27], [52],also providing additional features such as ease-of-access, ease-of-use, and reduced deployment costs [27].Newparadigms. The adoption of the CloudIoTparadigmenables newscenarios for smart services and applicationsbased on the extension of Cloud through the things [50], [52]: SaaS (Sensing as a Service) [50], [56], [27], providingubiquitous access to sensor data; SAaaS(Sensing and Actuation as a Service) [50],enabling automatic control logics implemented in theCloud; SEaaS(Sensor Event as aService) [50], [27], dis-patching messaging services triggered by sensorevents; SenaaS (Sensor as a Service) [56], enabling ubiquitousmanagement of remote sensors; DBaaS(DataBaseasaService)[56], enablingubiq-uitous database management; DaaS(DataasaService)[56], providingubiquitousaccess to any kind of data; EaaS(Ethernet asaService)[56], providingubiqui-tous layer-2 connectivity to remote devices; IPMaaS(IdentityandPolicyManagement asaSer-vice) [56], enablingubiquitousaccesstopolicyandidentity management functionalities; VSaaS(VideoSurveillanceas aService) [49], pro-viding ubiquitous access to recorded video and imple-menting complex analyses in the Cloud.III. APPLICATIONSIn this section we describe a wide set of applications thatare made possible or signicantly improved thanks to theFig. 3: Application scenarios driven by the CloudIoT paradigm.CloudIoTparadigm. For eachapplicationwe point out thechallenges, which we discuss in detail in Sec. IV.Healthcare. IoT and multimedia technologies have made theirentrance in the healthcare eld thanks to ambient-assistedlivingandtelemedicine[57]. Smartdevices, mobileInternet,and Cloud services contribute to the continuous and systematicinnovationof Healthcareandenablecost effective, efcient,timely, and high-quality ubiquitous medical services [41], [33].Pervasivehealthcareapplications generateavast amount ofsensor data that have to be managed properly for furtheranalysisandprocessing[29]. TheadoptionofCloudinthisscenario leads to the abstraction of technical details, eliminat-ingtheneedforexpertisein,orcontrolover,thetechnologyinfrastructure [15], [43], and it represents a promising solutionfor managing healthcare sensor data efciently [29]. It furthermakesmobiledevicessuitedforhealthinformationdelivery,access andcommunication, alsoonthego[46], enhancingmedical data security, availability, and redundancy [41], [15].Moreover, it enablestheexecution(intheCloud) of securemultimedia-based health services, overcoming the issue of run-ning heavy multimedia & security algorithms on devices withlimited computational capacity and small batteries [46]. In thiseld, commonissuesrelatedtomanagement, technology, se-curity, and law have been investigated: interoperability, systemsecurity, streaming Quality of Service (QoS), and dynamicallyincreasing storage are commonly considered obstacles [41].Smart City. IoT can provide a common middleware forfuture-oriented Smart-City services [17], [52], acquiring infor-mationfromdifferent heterogeneous sensinginfrastructures,accessing all kinds of geo-location and IoT technologies (e.g.,3Drepresentations throughRFIDsensors andgeo-tagging),andexposinginformationina uniformway. Anumber ofrecently proposed solutions suggest to use Cloud architecturesto enable the discovery, connection, and integration of sensorsandactuators, thuscreatingplatformsabletoprovisionandsupport ubiquitous connectivity and real-time applications forsmart cities [45]. Frameworks can consist of a sensor platform(withAPIsforsensingandactuating)andaCloudplatformfor theautomaticmanagement, analysis, andcontrol of bigdatafromlarge-scale, real worlddevices[52]. Thistypeofadvanced service model hides the complexity of the underlyingCloudinfrastructure, whilst at thesametimemeetingcom-plexpublicsector requirementsfor Cloud, suchassecurity,heterogeneity, interoperability, scalability, extensibility, highreactivity, and congurability [17], [52]. Moreover, Cloud-basedplatforms helptomake it easier for thirdparties todevelopandprovideIoTpluginsenablinganydevicetobeconnected to the Cloud [17]. While cities share commonconcernssuchastheneedtoeffectivelyshareinformationwithinandbetweencitiesandthedesireforenhancedcross-border protocols they lack a common infrastructure andmethodologyfor collaborating[17], whichmayresult fromthe application of a holistic approach [52]. Common issues arerelated to security, resilience, and real-time interactions [52].Smart HomeandSmart Metering. IoThaslargeapplica-tioninhomeenvironments, whereheterogeneousembeddeddevices enable the automation of common in-house activities.Inthisscenario, theCloudisthebestcandidateforbuildingexibleapplicationswithonlyafewlinesof code, makinghomeautomationatrivialtask[39]. Inordertoletavarietyof independent single-family smart homes access reusableservices over the Internet, the resulting solution should satisfythree crucial requirements [54]: internal network intercon-nection(i.e., everydigital appliance insmart home shouldbe able to interconnect with any other), intelligent remotecontrol (i.e., appliances and services in the smart home shouldbeintelligentlymanageablebyanydevicefromanywhere),and automation (i.e., interconnected appliances within thehome should implement their functions via linking to servicesprovided by smart-home oriented Cloud). Several smart-homeapplications proposedinliteratureinvolve(wireless) sensornetworksandimplementsmartmeteringsolutionstoproviderecognition of appliances [24], intelligent management ofenergy consumption [24], lighting, heating, and air condi-tioning[36]. Several issuesmust beresolvedtomaterializethisvision[36]. Homedevicesshouldbeweb-enabled, theirdiscovery and service description should be standardized, andtheinteractionwiththemshouldbeuniform. Administrationand control of devices could be leveraged by deploying morepowerful computingdevices, actingasmediatorsamongIoTdevices andCloudcomponents, for implementingcomplexfunctionalities on top of them, mitigating the volume and thefrequency of communications with the Cloud.Video Surveillance. Intelligent video surveillance has becomeatool ofthegreatest importanceforseveral security-relatedapplications. As an alternative to in-house, self-contained man-agement systems, complex video analytics require Cloud-basedsolutions (VSaaS [49]) to properly satisfy the requirements ofstorage (e.g., stored media is centrally secured, fault-tolerant,on-demand, scalable, andaccessibleat high-speed)andpro-cessing(e.g., videoprocessing, computer visionalgorithmsandpatternrecognitionmodules toextract knowledgefromscenes). Proposed solutions [34] intelligently store and managevideocontentoriginatingfrom(IPandanalog)cameras, andefciently deliver it to multiple user devices through theInternet, by distributing the processing tasks over the physicalserver resources on-demand, in a load-balanced and fault-tolerant fashion.Automotive and Smart Mobility. As an emerging technology,IoTis expected to offer promising solutions to transformtransportationsystems andautomobileservices (i.e., Intelli-gent TransportationSystems, ITS). TheintegrationofCloudtechnologies with WSNs, RFID, satellite networks, and otherintelligent transportationtechnologiesrepresentsapromisingopportunity to tackle the main current challenges [38]. Anewgenerationof IoT-basedvehicular data Clouds canbedeveloped and deployed to bring many business benets, suchas increasing road safety, reducing road congestion, managingtrafc, and recommending car maintenance or xing [38]. Theliteratureproposesseveralexamplesofmulti-layered, Cloud-basedvehicular dataplatformsthat mergeCloudcomputingand IoT technologies. These platforms aim at providing real-time, cheap, secure, andon-demandservices tocustomers,through different types of Clouds, which also include tempo-rary vehicular Clouds (i.e., formed by the vehicles representingtheClouddatacenters [20]). Vehicular Clouds aredesignedtoexpandtheconventional Cloudsinorder toincreaseon-demandthe whole Cloudcomputing, processing, andstor-age capabilities, by using under-utilized facilities of vehicles.Thankstotheseplatforms, innovative, vehicular, dataCloudservices can be deployed (e.g., intelligent parking Cloudservice, or vehicular data-miningCloudservice) [38]. Moreingeneral, ethernet andIP-basedrouting(beinglessexpen-siveandmoreexiblethanrelatedtechnologies)areclaimedtobeveryimportant technologiesfor futurecommunicationnetworksinelectricvehicles, enablingthelinkbetweenthevehicle electronics andthe Internet, integratingthe vehicleinto a typical IoT, and meeting the demand for powerfulcommunication with Cloud-based services [37]. Several issueshave been identied in literature affecting this applicationscenario[38], [20]. Thehugenumber of vehicles andtheirdynamically changing number make system scalability difcultto achieve. Vehicles moving at various speeds frequently causeintermittent communication impacting performance, reliability,and QoS. The lack of an established infrastructure makes it dif-culttoimplementeffectiveauthenticationandauthorizationmechanisms, withimpactsonsecurityandprivacyprovision[44]. The lack of global standards and experimental studies onrealistic ITS-Clouds affect interoperability.Smart Energy and Smart Grid. IoTand Cloud can beeffectivelymergedtoprovideintelligent management ofen-ergy distribution and consumption in both local and wide areaheterogeneous environments. Inthe rst case, for instance,lighting could be provided where and when strictly necessarybyexploitingtheinformationcollectedbydifferent typesofnodes [36]. Such nodes have sensing, processing, and network-ing capabilities, but limited resources. Hence, computing tasksshould be properly distributed among them or demanded to theCloud, where more complex and comprehensive decisions canbe made. In the second case, the problem on energy alternativeand compatible use can be solved by integrating system data inthe Cloud, while providing self-healing, mutual operation andparticipation of the users, optimal electricity quality, distributedgeneration, and demand response [55]. The integration ofCloudplatformsinthisIoTscenarioincreasestheconcernsabout security and privacy issues for Smart Grid softwaredeployment for utilities. These concerns should be adequatelyaddressed to realize the full potential of such application:consumers should gain more condence in sharing data to helpimproving and optimizing services offered [51].IV. HOT TOPICS AND RELATED CHALLENGESMigrating IoTapplication specic data into the Cloudoffersgreatconvenience,suchasreductionofcostandcom-plexityrelatedtodirect hardwaremanagement. Accordingly,the complex scenario of CloudIoTincludes many aspectsrelated to several heterogeneous topics, each of themim-posingchallengeswhenrequiringspeciccapabilitiestobesatised.Forinstance,thefollowingcapabilitiesarerequiredtoguaranteetrustedandefcient services: security, privacy,reliability, availability, portability, and semantic interoperabil-ity[19], [30]. Whencritical IoTapplicationsmovetowardsthe Cloud, newconcerns arise due to the lack of someessential properties: trust intheserviceprovider, knowledgeabout service level agreements (SLAs), and knowledge aboutthephysical locationof thedatastorage. Accordingly, newchallenges require specic attention [19]: the distributed natureof the system exposes several signicant attacks (e.g., sessionriding, SQLinjection, crosssitescripting, andside-channel)and vulnerabilities (e.g., session hijacking and virtual machineescape); multi-tenancycouldcompromisesecurityandmaylead to sensitive information leakage; public key cryptographycannot beappliedat all layersduetothecomputingpowerconstraints imposed by the things. As summarized in Table II,inthefollowingwereporthowprovidingcertaincapabilitieshas proved challenging in the recent literature when addressingseveral main topics pertaining to CloudIoT.Servicedelivery. IoTservicesaretypicallyprovidedasiso-latedvertical solutions, inwhichall systemcomponentsaretightlycoupledtothespecicapplicationcontext. For eachdeployedapplicationservice, providers have tosurveytar-get application environments, analyze its requirements, selecthardware devices, integrate heterogeneous subsystems, developapplications, provide computing infrastructure, and provideservice maintenance. On the other hand, leveraging Cloud ser-vice delivery models, CloudIoT can enable efcient, scalable,and easily-extensible IoT service delivery [42]. Although PaaS-like models would represent a generic solution for facilitatingthe deployment of most IoT applications, their adoption movessomeIoTchallengesintotheCloudworld. Forinstance, theinteraction with (management of) huge amounts of highly het-erogeneous things (produced data) has to be properly addressedinto the Cloud at different levels.NetworkingandCommunicationProtocols. CloudIoTin-volves machine-to-machine (M2M) communications amongmany heterogeneous devices with different protocols [19],whichdependonthespecicapplicationscenario. Dealingwiththis heterogeneitytomanage the things ina uniformfashionwhile providingrequiredperformance [22], [18] ischallenging. Themajorityof applicationareas does not in-volve mobility: in stationary scenarios, IoT often adopts IEEE802.15.4/6LoWPANsolutions[48]. Ontheotherhand, otherscenarios such as vehicular networks mostly adopt IEEE802.11pspecications. However, beingWiFi andBluetooththe most widely used radio technologies for personal networks,their adoption for IoT applications is increasing: they representa cheaper solution, most mobile devices already support them(e.g., smart phones), andbothstandardsarebecomingmoreandmorelowpower. Insomeothercases, whenpowercon-straints are less critical, GPRS is used for Internet connectivity,but it results in a very costly solution (e.g., multiple SIM cardsare necessary) [48].Bigdata. Withanestimatednumber of 50billiondevicesthat will benetworkedby2020, specicattentionmust bepaidtotransportation, storage, access, andprocessingofthehuge amount of data they will produce. Thanks to the recentdevelopment intechnologies, IoTwill be one of the mainsources of big data, and Cloud will enable to store it for longtime andtoperformcomplexanalyses onit. The ubiquityof mobiledevicesandsensor pervasiveness, indeedcall forscalable computing platforms (every day 2.5 quintillion bytesofdataarecreated)[28]. Handlingthisdataconvenientlyisa critical challenge, as the overall applicationperformanceis highlydependent onthe properties of the data manage-ment service[28]. Hence, followingtheNoSQLmovement,bothcommercialandopensourcesolutionsadoptalternativedatabase technologies for big data [25]: time-series, key-value,document store, wide column stores, and graph databases.Unfortunately, no perfect data management solution exists forthe Cloud to manage big data [56].Sensor Networks. Sensor networks have been dened as themajor enabler of IoT [56] and as one of the ve technologiesthat will shape the world, offering the ability to measure, infer,and understand environmental indicators, from delicate ecolo-gies and natural resources to urban environments [35]. Recenttechnological advances have made efcient, low-cost, and low-power miniaturizeddevices availablefor useinlarge-scale,remote sensing applications [14]. Moreover, smartphones, eventhoughlimitedbypower consumptionandreliability, comewith a variety of sensors (GPS, accelerometer, digital compass,microphone, andcamera), enablingawiderangeof mobileapplicationsindifferent domainsofIoT. Inthiscontext, thetimely processing of huge and streaming sensor data, subjectto energy and network constraints and uncertainties, has beenidentiedas the mainchallenge [58]. Cloudprovides newopportunities inaggregatingsensor data andexploitingtheaggregates for larger coverage and relevancy, but at the sametime affects privacy and security [58]. Furthermore, beinglackof mobilitya typical aspect of commonIoTdevices,the mobility of sensors introduced by smartphones as well aswearable electronics represents a new challenge [48].User Participation. The investment into omnipresent Internet-capable devices is not reasonable in every scenario. Therefore,it wouldbe convenient togive the opportunitytousers toparticipate in submitting data that represent a thing [16].Users could also be empowered with new building blocks andtools: accelerators, frameworks, andtoolkitswill enabletheparticipationofusersinIoTasdoneintheInternet throughWikis and Blogs [26]. Such tools and techniques should enableresearchers and design professionals to learn about users work,givingusersanactiveroleintechnologydesign. Toachievethis, these tools should allow users to easily experiment variousdesignpossibilitiesinacost-effectiveway. Inthisscenario,one challenge is toprovide adequate prototypingtools forimplementing a cooperative prototyping approach, where usersand designers explore together applications and their relations.Monitoring. As largelydocumentedintheliterature, mon-itoring is an essential activity in Cloud environments forcapacity planning, for managing resources, SLAs, performanceand security, and for troubleshooting [13]. As a consequence,CloudIoTinherits the same monitoring requirements fromCloud, but the related challenges are further affected byvolume, variety, and velocity characteristics of IoT.V. PLATFORMS, SERVICES, AND RESEARCHPROJECTSInthissectionwedescribeopensourceandcommercialplatforms for CloudIoT and three research projects focused onthistopic, whichwebelieveareanexampleoftheresearchefforts funded in this important area.A. PlatformsThereareseveral opensourceandcommercial platformsfor CloudandIoTintegration. Most of themareaimedatsolvingoneofthemainissuesinthiseldthatisrelatedtothe heterogeneity of things and Clouds. These platforms try tobridge this gap implementing a middleware towards the thingsandanother towards the Cloud, andtheytypicallyprovideanAPItowardstheapplications. Otherplatformsareinsteadbound to specic hardware devices or Clouds. Tab. III reportsthe main characteristics of these platforms and services. In thefollowing we review both types of platforms, starting with theones belonging to the former group.IoTCloud [1] is an open source project aimed at integratingthe things (smart phones, tablets, robots, Web pages, etc.)withbackendsformanagingsensorsandtheirmessagesandforprovidinganAPItoapplicationsinterestedinthesedata.The platform has been showcased with video sensors (i.e., IPcameras) on the FutureGrid Cloud testbed [31]. The software isavailable online through github. OpenIoT [2] is another opensourceeffort fosteredbyaresearchproject nancedbytheEU. The project aims at providing a middleware to congureanddeployalgorithms for collectionandlteringmessagesbythings, whileat thesametimegeneratingandprocessingevents for interested applications. Among the main focuses ofOpenIoTarethemobilityaspectsofIoTforenergy-efcientorchestration of data collecting and transmission to the Cloud.The project website [2] contains different videos showingpossibleapplicationsinseveral scenarios. Alsothissoftwareis available online through github. There are also projectsspecically aimed at creating toolkits for the interaction of IoTand Cloud. For example, IoT Toolkit (run by a Silicon ValleybasedorganizationcalledOSIOT) [3] aims at developingatoolkit that allows to glue the several protocols available bothfor the things, for the Cloud, and for the applications.Asforthecommercial solutions(e.g. platformstypicallyboundtospecicthingsor Clouds), Postscapespublishesacatalogofprojects, events, interviews, andcompany/joblist-ings within the industry [4]. Interesting examples include opensourceprojectsrunbyprivatecompanies. For example, theopen source project of NimBits [5] provides a set of softwaretobeinstalledonprivateor publicClouds (mainlyGoogleAppEngine)tocreateaPaaSthat collectsdatafromthingsandtriggerscomputationsoralertswhenspecicconditionsareveried. ThecompanyprovidesalsoaCloudserviceforrunning this software that is free of charge but has someusage limitations. On the other side, there are companies thatprovide things ready to be integrated in Clouds. For example,openPicus [6] is an Italian company which builds things (e.g.TABLE II: Topics and challenges pertaining CloudIoT.EfciencyScalabilityExtensibilityReliabilitySecurityPrivacyInteroperabilityAvailabilityMobilityPervasivenessEnergysavingUniformityMaintainabilityCosteffectivenessLargescaleService deliveryMonitoring Big DataSensor Networks User Participation Networking and Communication Protocols small sensors equipped with WiFi or GPRS connectivity) usinganopenhardwareapproach. Theideaistobuildverycheapproducts that have full TCP/IP stack implemented and HTTPserver on-board, whichallows tointeract withthemusingsimple RESTful APIs.Finally, there are also several services (es. Xively [7],Open.Sen.se[8], ThingSpeak[9]) that allowtocollect datafrom things and to store these data on the Cloud offered by theservice provider. These services typically provide an API anddifferent example applications to use the data collected by thethings. Startingfromtheseservices, companieshavecreatedtoolkits for their integration in CloudIoTframeworks. Forexample, NetLab [10] is a toolkit for interaction among phys-ical and digital objects (e.g. controlling video movies througharduino). NetLab has created two widgets called CouldIn andCloudOut that allow to interact with several CloudIoT services.In particular, they allow to periodically send data from thingstotheseservicesor toperiodicallyretrievedatafromtheseservices. Compliant services include Xively (formerly COSMand Pachube), Open.Sen.se, and ThingSpeak.TABLE III: Platforms, services and research projects.ProprietarythingsOpenthingsPrivatecloudPubliccloudFreeOpensourceApplicationAPILastupdateReadytouseIoTCloud n/a n/aMay13 OpenIoTApr14 IoT Toolkit n/a n/a n/a Dec 13NimBitspartly n/a Mar 14 openPicusn/a n/a n/a n/an/a n/aXivelyn/a Open.Sen.sen/a ThingSpeakn/a NetLabn/a Dec 13 B. Research ProjectsClouT[11], which stands for Cloud of things, is aresearchproject runbyindustrial andresearchpartners aswell as city administrations fromEurope and Japan. Thepartnersaimat developinginfrastructure, services, toolsandapplications for municipalities and their stakeholders (citizens,servicedevelopers, etc.)tocreate, deploy, andmanageuser-centric applications based on IoT and Cloud integration. Targetapplications include enhanced public transportation, increasedcitizen participation through mobile devices (e.g. to photographand record situations of interest to city administrators), safetymanagement, cityeventmonitoring, andemergencymanage-ment.IoT6 is a European research project on the future Internetof Things. It aims at exploiting IPv6 and related standards (e.g.,6LoWPAN, CORE, COAP) to overcome current shortcomings(e.g. in terms of fragmentation) in the area of IoT research anddevelopment.Itsmainobjectivesaretoresearch,design,anddevelop a highly scalable IPv6-based Service-Oriented Archi-tecture to achieve interoperability, mobility, Cloud computingintegration, and intelligence distribution among heterogeneousthings, applications, and services.The OpenIoTproject, cited before for its open sourceplatform, aims at creating an open source middleware forgetting information fromheterogeneous things, hiding thedifferences among these objects. The project explores efcientwaystouseandmanageCloudenvironmentsforthingsandresources(suchassensors,actuators,andsmartdevices)andoffering utility-based (i.e., pay-as-you-go) IoT services.VI. OPEN ISSUES AND FUTURE DIRECTIONSThankstotheanalyseswehavedoneinSec. II, inSec.III, and in Sec. IV, here we resume the main issues related toCloudIoTstill requiringresearcheffortsandpoint out somefuture directions.The need for Standards. Even though the scientic commu-nitygavemultiplecontributionstothedeploymentandstan-dardization of IoT and Cloud paradigms, a clear necessity ofstandard protocols, architectures and APIs is being demandedin order to facilitate the interconnection among heterogeneoussmart objects andthecreationof enhancedservices, whichrealizetheCloudIoTparadigm[52]. InparticularMobile-to-Mobile (M2M) is a leading paradigm, but there is a very littlestandardization for it. Hence the several existing solutions usestandardInternet, cellular, andWebtechnologies. Moreover,the state of the art lacks domain specic environments for rapiddevelopmentandefcientCloudIoTservicedelivery. Indeed,most architecturesproposedat theinitial stageofIoTeitherhave come from the WSN perspective or are based on Cloudat the center.EnergyEfcient Sensing. WSNtypicallyconsists of low-cost, low-power, and energy-constrained sensors. Each opera-tion, calculation, and inter-communication consumes the nodeenergy. Also when specic sensors are replaced by smartdevices, energysavingisamajor issueandenergyefcientmanagement ismorethandesirable. Several techniquescanbeadoptedtoachieveenergysaving: forinstance, compres-sive sensing enables reduced signal measurements withoutimpactingaccuratereconstructionof thesignal andutilizessynchronous communication to reduce the transmitting powerof each sensor. On the other side, CloudIoT, involving Cloudparadigm is not affected by energy issues. Indeed, it is possibleto lease on-demand computing and storage resources throughCloud in an optimal manner from the energetic point of view.In addition, local computations can be shifted to the Cloud, inorder to save computational resources and device energy.NewProtocols. MACandroutingprotocolsarecritical forthe system to work efciently. Even though several protocolshave been proposed for various domains (with TDMA, CSMA,andFDMAschemes), none of themhas beenacceptedasastandard, andwiththegrowingnumber of things, furtherresearch is required. Energy is the main consideration for rout-ing protocols. Although the literature reports existing routingprotocols can work with minor modications in IoT scenarios(considering that the number of hops in multi-hop scenario islimited and that a backbone is typically available) [35], IETFROLLworkgroup(whichdealswithroutingoverlow-powerlossy networks) claims that existing routing protocols such asOSPF, IS-IS, AODV, and OLSR have been found to not satisfyLowPower and Lossy Networks (LLNs) specic routingrequirements in their current form (e.g. optimization for energysaving, typical point-to-multipoint trafc patterns, restrictedframe-sizes, limits in trading efciency for generality). HenceRPL (ripple) protocol draft has been proposed [12].Participative Sensing. The problemof missingsamples iscritical in people-centric sensing. Indeed, relying on usersvolunteering data is a severe limitation to the ability toproducemeaningful data. Moreoverdataownershipandpri-vacyissuesmust beaddressedandappropriateparticipationincentives must be identied to achieve genuine end-userengagement [35].Complex Data Mining. Current technologies are not able tocompletelysolveall issuesrelatedtothecomplexityof bigdata. The percentage of data an organization can analyse is onthe decline: due to the high number of big data producers andthehighfrequencyofdatageneration, thegapbetweendataavailable to organizations and data they can process is gettingwider.Approximateresultsaretypicallyordersofmagnitudefaster compared to traditional query execution [56]. Researcheffortsarerequiredtomeet thechallengeof bigdata. Newtechnologiesandqueryoptimizationsarerequiredtoanalyzelarge volumes of data faster with efcient resource and powerconsumption [56]. Big data consists of high-valued data mixedwithdirtyanderroneousdata. Extractinguseful informationat different spatial andtemporal resolutionsisachallengingprobleminarticial intelligenceresearch. Whilestate-of-artmethods use shallowlearning, an emerging focus is deeplearning, which aims at learning multiple levels of abstraction,which can be used to interpret given data. Heterogeneousspatio-temporal (geo-related and sparsely distributed) datacoming from IoT elaborations are not always ready for directconsumption using visualization platforms. New visualizationschemes have to be developed (e.g. 3D, GIS), in order to createattractive and easy to understand visualizations [35].Cloud Capabilities. Security, which concerns every networkedenvironment, is a major issue for CloudIoT. Indeed both its IoTside (i.e., RFID, WSN) and its Cloud side are vulnerable to anumber of attacks. In IoT context, encryption can ensure datacondentiality, integrity, andauthenticity. Howeverit cannotaddress insider attacks (e.g. during WSNreprogramming)andit is difcult toimplement oncomputation-constraineddevices. RFIDis themost vulnerablecomponent, sincenohigher level of intelligence can be enabled on it. Also, securityaspects related to Cloud require attention since Cloud handlestheeconomics, alongwithdataandtools. Moreover, Cloudplatformsneedtobeenhancedtosupport therapidcreationof applications, byprovidingdomainspecic programmingtools and environments and seamless execution of applications,harnessing capabilities of multiple dynamic and heterogeneousresources, tomeetQoSrequirementsofdiverseusers. Cloudschedulingalgorithms needtomanage taskduplicationforfailuremanagement, inordertodeliverservicesinareliablemanner. Moreover, they should be able to deal with QoSparameters.FogComputing. Fogcomputingis anextensionof classicCloud computing to the edge of the network (as fog is a cloudclose to the ground). It has been designed to support IoT appli-cations, characterizedbylatencyconstraintsandrequirementfor mobility and geo-distribution [21]. Even though computing,storageandnetworkingareresourcesofboththeCloudandtheFog, thelatter hasspeciccharacteristics: edgelocationandlocationawareness implyinglowlatency; geographicaldistribution and a very large number of nodes in contrasttocentralizedCloud; support for mobility(throughwirelessaccess) and real-time interaction (instead of batch processes);support for interplay with the Cloud.ACKNOWLEDGEMENTSThisworkispartiallyfundedbytheMIURprojects:PLATINO(PON0101007), SMARTHEALTH(PON04a2C) andSIRIO(PON0102425).REFERENCES[1] https://sites.google.com/site/opensourceiotcloud/.[2] http://www.openiot.eu/.[3] http://iot-toolkit.com/.[4] http://postscapes.com/internet-of-things-cloud.[5] http://www.nimbits.com/.[6] http://www.iot-a.eu/.[7] https://xively.com/.[8] http://open.sen.se/.[9] https://thingspeak.com/.[10] http://www.netlabtoolkit.org/learning/tutorials/iot-cloud-services/.[11] http://clout-project.eu/home2/.[12] https://datatracker.ietf.org/wg/roll/charter/.[13] G. Aceto, A. Botta, W. De Donato, and A. Pescap` e. Cloud monitoring:A survey. Computer Networks, 57(9):20932115, 2013.[14] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci. Wirelesssensor networks: a survey. Computer networks, 2002.[15] F. Alagozetal. FromcloudcomputingtomobileInternet, fromuserfocus to culture and hedonism: the crucible of mobile health care andwellness applications. In ICPCA 2010. IEEE, 2010.[16] C. Atkins et al. A Cloud Service for End-User Participation ConcerningtheInternetofThings. InSignal-ImageTechnology&Internet-BasedSystems (SITIS), 2013 International Conference on. IEEE, 2013.[17] P. Ballon, J. Glidden, P. Kranas, A. Menychtas, S. Ruston, and S. VanDer Graaf. Is there a need for a cloud platformfor europeansmart cities? IneChallenges e-2011ConferenceProceedings, IIMCInternational Information Management Corporation, 2011.[18] M. Bernaschi, F. Cacace, A. Pescape, and S. Za. Analysis andexperimentation over heterogeneous wireless networks. In Tridentcom.IEEE, 2005.[19] T. Bhattasali, R. Chaki, and N. Chaki. Department of ComputerScience & Engineering, University of Calcutta, Kolkata, India. In IndiaConference (INDICON), 2013 Annual IEEE, pages 16. IEEE, 2013.[20] S. Bitam and A. Mellouk. ITS-cloud: Cloud computing for Intelligenttransportation system. In Global Communications Conference (GLOBE-COM), 2012 IEEE, pages 20542059. IEEE, 2012.[21] F. Bonomi, R. Milito, J. Zhu, and S. Addepalli. Fog computing and itsrole in the internet of things. In Proceedings of the rst edition of theMCC workshop on Mobile cloud computing, pages 1316. ACM, 2012.[22] A. Botta, A. Pescap e, and G. Ventre. Quality of service statistics overheterogeneous networks: Analysis and applications. European Journalof Operational Research, 191(3):10751088, 2008.[23] H.-C. Chao. Internet of things and cloud computing for future internet.In Ubiquitous Intelligence and Computing, Lecture Notes in ComputerScience. 2011.[24] S.-Y. Chen, C.-F. Lai, Y.-M. Huang, and Y.-L. Jeng. Intelligenthome-appliance recognition over IoT cloud network. In WirelessCommunicationsandMobileComputingConference(IWCMC), 20139th International, pages 639643. IEEE, 2013.[25] A. Copie, T.-F. Fortis, and V. I. Munteanu. Benchmarking CloudDatabases for theRequirements of theInternet of Things. In34thInternational Conference on Information Technology Interfaces, ITI2013, pages 7782, 2013.[26] I. P. Cvijikj andF. Michahelles. TheToolkit ApproachforEnd-userParticipationintheInternetofThings. InArchitectingtheInternetofThings, pages 6596. Springer, 2011.[27] S. K. Dash, S. Mohapatra, and P. K. Pattnaik. A Survey on Applicationof WirelessSensor NetworkUsingCloudComputing. InternationalJournal of Computer science &EngineeringTechnologies (E-ISSN:2044-6004), 1(4):5055, 2010.[28] C. Dobre and F. Xhafa. Intelligent services for Big Data science. FutureGeneration Computer Systems, 2013.[29] C. Doukas andI. Maglogiannis. Bringingiot andcloudcomputingtowards pervasive healthcare. In Innovative Mobile and InternetServices in Ubiquitous Computing (IMIS), 2012 Sixth InternationalConference on, pages 922926. IEEE, 2012.[30] European Commission. Denition of a research and innovation policyleveragingCloudComputingandIoTcombination. Tenderspecica-tions, SMART 2013/0037, 2013.[31] G. Fox, G. von Laszewski, J. Diaz, K. Keahey, J. Fortes, R. Figueiredo,S. Smallen, W. Smith, and A. Grimshaw. Futuregrid A recongurabletestbedforcloud, hpcandgridcomputing. ContemporaryHighPer-formance Computing: From Petascale toward Exascale, ComputationalScience. Chapman and Hall/CRC, 2013.[32] G. C. Fox, S. Kamburugamuve, andR. D. Hartman. Architectureandmeasuredcharacteristicsof acloudbasedinternet of things. InCollaboration Technologies and Systems (CTS), 2012 InternationalConference on, pages 612. IEEE, 2012.[33] D. Gachet, M. deBuenaga, F. Aparicio, andV. Padr on. Integratinginternet of things and cloud computing for health services provisioning:The virtual cloudcarer project. InInnovative Mobile andInternetServices in Ubiquitous Computing (IMIS), 2012 Sixth InternationalConference on, pages 918921. IEEE, 2012.[34] F. Gao. VSaaSModel onDRAGON-Lab. International Journal ofMultimedia & Ubiquitous Engineering, 8(4), 2013.[35] J. Gubbi, R. Buyya, S. Marusic, and M. Palaniswami. Internet of Things(IoT): Avision, architectural elements, andfuturedirections. FutureGeneration Computer Systems, 29(7):16451660, 2013.[36] D.-M. HanandJ.-H. Lim. Smart homeenergymanagement systemusing IEEE 802.15. 4 and zigbee. Consumer Electronics, IEEETransactions on, 56(3):14031410, 2010.[37] P. Hank, S. M uller, O. Vermesan, and J. Van Den Keybus. Automotiveethernet: in-vehiclenetworkingandsmart mobility. InProceedingsof theConferenceonDesign, AutomationandTest inEurope, pages17351739. EDA Consortium, 2013.[38] W. He, G. Yan, and L. Xu. Developing Vehicular Data Cloud Servicesin the IoT Environment. 2009.[39] A. Kamilaris et al. The smart home meets the web of things.International Journal of Ad Hoc and Ubiquitous Computing, 2011.[40] B. Kitchenham. Procedures for performing systematic reviews. Keele,UK, Keele University, 33:2004, 2004.[41] A. M.-H. Kuo. Opportunitiesandchallengesof cloudcomputingtoimprovehealthcareservices. Journal of medical Internet research,13(3), 2011.[42] F. Li, M. V ogler, M. Claeens, and S. Dustdar. Efcient and scalableIoTservicedeliveryonCloud. InCloudComputing(CLOUD), 2013IEEE Sixth International Conference on, pages 740747. IEEE, 2013.[43] H. L ohr, A.-R. Sadeghi, and M. Winandy. Securing the e-healthcloud. In Proceedings of the 1st ACM International Health InformaticsSymposium, pages 220229. ACM, 2010.[44] P. Marchetta, E. Natale, A. Salvi, A. Tirri, M. Tufo, and D. De Pasquale.Trusted information and security in smart mobility scenarios: The caseof s2-move project. In Algorithms and Architectures for ParallelProcessing, pages 185192. Springer, 2013.[45] N. Mitton, S. Papavassiliou, A. Puliato, and K. S. Trivedi.CombiningCloud and sensors in a smart city environment. EURASIP Journal onWireless Communications and Networking, 2012(1):110, 2012.[46] M. Nkosi and F. Mekuria. Cloud computing for enhanced mobile healthapplications. In Cloud Computing Technology and Science (CloudCom),2010 IEEE Second International Conference on. IEEE, 2010.[47] P. Parwekar.From Internet of Things towards cloud of things.In Com-puter and Communication Technology (ICCCT), 2011 2nd InternationalConference on, pages 329333. IEEE, 2011.[48] P. P. Pereira, J. Eliasson, R. Kyusakov, J. Delsing, A. Raayatinezhad,and M. Johansson. Enabling cloud connectivity for mobile internet ofthingsapplications. InServiceOrientedSystemEngineering(SOSE),2013 IEEE 7th International Symposium on. IEEE, 2013.[49] A. Prati, R. Vezzani, M. Fornaciari, and R. Cucchiara. Intelligent VideoSurveillance as a Service. In Intelligent Multimedia Surveillance, pages116. Springer, 2013.[50] B. P. Rao, P. Saluia, N. Sharma, A. Mittal, andS. V. Sharma. Cloudcomputingfor Internet of Things &sensingbasedapplications. InSensingTechnology(ICST), 2012SixthInternational Conferenceon,pages 374380. IEEE, 2012.[51] Y. Simmhan, A. G. Kumbhare, B. Cao, and V. Prasanna. An analysisof security and privacy issues in smart grid software architecturesonclouds. InCloudComputing(CLOUD), 2011IEEEInternationalConference on, pages 582589. IEEE, 2011.[52] G. Suciu, A. Vulpe, S. Halunga, O. Fratu, G. Todoran, andV. Suciu.SmartCitiesBuiltonResilientCloudComputingandSecureInternetof Things. InControl SystemsandComputerScience(CSCS), 201319th International Conference on, pages 513518. IEEE, 2013.[53] D. Yao, C. Yu, H. Jin, and J. Zhou. Energy Efcient Task Scheduling inMobile Cloud Computing. In Network and Parallel Computing, pages344355. Springer, 2013.[54] X. YeandJ. Huang. AframeworkforCloud-basedSmartHome. InComputerScienceandNetworkTechnology(ICCSNT), 2011Interna-tional Conference on, volume 2, pages 894897. IEEE, 2011.[55] M. Yun and B. Yuxin. Research on the architecture and key technologyof Internet of Things (IoT) applied on smart grid. In Advances in EnergyEngineering (ICAEE), 2010 International Conference on, pages 6972.IEEE, 2010.[56] A. Zaslavsky, C. Perera, and D. Georgakopoulos. Sensing as a serviceand big data. arXiv preprint arXiv:1301.0159, 2013.[57] Q. Zhang, L. Cheng, and R. Boutaba. Cloud computing: state-of-the-artand research challenges. Journal of internet services and applications,1(1):718, 2010.[58] F. Zhao. Sensorsmeet thecloud: Planetary-scaledistributedsensingand decision making. In Cognitive Informatics (ICCI), 2010 9th IEEEInternational Conference on, pages 998998. IEEE, 2010.[59] J. Zhou, T. Leppanen, E. Harjula, M. Ylianttila, T. Ojala, C. Yu, andH. Jin. Cloudthings: A common architecture for integrating the internetof things with cloud computing. In CSCWD, 2013. IEEE.[60] P.Zikopoulos,C.Eaton,etal. Understandingbigdata:Analyticsforenterpriseclasshadoopandstreamingdata. McGraw-Hill OsborneMedia, 2011.