icirrus contract no. 644526 1 jan 2015 – 31 dec 2017 · enb enhanced node b (radio base station...
TRANSCRIPT
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
This project has received funding from the European Union’s Horizon 2020
research and innovation programme under grant agreement No 644526
intelligent Converged network consolidating Radio and optical access aRound USer equipment
DELIVERABLE: D5.4
Validation test results and analysis evaluation
Contract number: 644526
Project acronym: iCIRRUS
Project title: Intelligent converged network consolidating radio and optical access around user equipment
Project duration: 1 January 2015 – 31 December 2017
Coordinator: Nathan Gomes, University of Kent, Canterbury, UK
Deliverable Number: D5.4
Type: Report
Dissemination level Public
Date submitted: 20 February 2018
Editors: Howard Thomas (VIAVI)
Authors / contributors (contributing partners)
Kai Habel (HHI), Jim Zou (ADVA), Philippos Assimakopoulos, Nathan Gomes, Huiling Zhu, Yuan Kai (UniKent), Howard Thomas (VIAVI), Xuan Du (UEssex), Philippe Chanclou (Orange), Patrik Ritoša (TS), Guadalupe Flores (WT), Gregor Linne (IAF)
Internal reviewers Mike Parker (UEssex), Michael Georgiades (PrimeTel), Nathan Gomes (UniKent)
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 2 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Documenthistory
0.0 Document creation 17/10/2017
0.1 Updated TOC to reflect D5.3 included Section on aggregation of results and analysis
30/11/2017
0.2 Included inputs to Validation test from ADVA, HHI. Elaboration of KPI Section by VIAVI
21/12/2017
0.25 Demo schematic update and added LTE femto‐cell case description 22/12/2017
0.3 Wellness version
0.4 Howard comments 3/01/2018
0.45 Minor updates and accept additions 4/01/2018
0.5 Initial inputs about mobile cloud from UEssex 07/01/2018
0.55 Combine V5, V7 & Orange 15/01/2018
0.6 Integrate Philippos content 15/01/2018
0.65 Accept Xuan content 17/01/2018
0.7 Changes accepted, acronym list, Summary text from Jim 17/01/2018
0.72 Additions from Gregor on CPRI 22/01/2018
0.75 Accept changes, remove fixed comments, minor edits 24/01/2018
0.8 1st attempt at completing intro and the evaluation Sections 24/01/2018
0.85 Re‐emphasis, re‐write in context of 5G KPIs 09/02/2018
0.90 Accept changes to format, typos, include text from Kai and my editspost review. This is a baseline to edit the Demo as a KPI in a broader scheme
12/02/2018
0.91 Restructure and emphasise KPI validation 13/02/2018
0.92 revision of text 3.9.1 13/02/2018
0.93 Re‐structure to separate testbed KPI from demo KPI 14/02/2018
0.94 Added evaluation of the Showcase section, incorporated inputs on D2D and mobile cloud
17/02/2018
1.00 Added Conclusions and general tidy‐up 19/02/2018
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 3 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Abstract
The deliverable presents the results obtained throughout the iCIRRUS project and compares them
with the 5G requirements including business, societal and performance KPIs. One KPI that is relevant
to both the business and societal KPIs is the number of proof of concept demonstrators as this is
recognised as a measure of “European availability of a competitive industrial offer for 5G services
and technologies.” Consequently, the showcase demonstrator hosted by Telekom Slovenije in
December 2017 is a focus of this report as is the separate demonstrator for evolved‐fronthaul higher
layer split hosted by Orange. Three families of use‐case are explored: enhanced front haul, mobile
cloud, and network controlled device‐to‐device communication, aiming to demonstrate that the
planned 5G networking requirements are met. The architecture and component parts of the
demonstrator are described and the performances of the use‐cases are presented. The results are
encouraging, in that all the iCIRRUS quantitative performance targets are met and there are
significant contributions towards the 5G performance KPIs. Additionally, significant contributions
towards achievement of the 5G business and societal are demonstrated by the three use‐case
classes.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 4 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
ExecutiveSummary
The report presents the results obtained throughout the iCIRRUS project and, to validate them
against 5G Key Performance Indicators (KPIs), compares the results against 5G requirements
including business, societal and performance KPIs as summarised in [1].
Two elements of elements of the KPI validation relate to technology demonstrators. One, hosted by
Orange, addressed evolved‐fronthaul with a higher layer split. The second, hosted by Telekom
Slovenije in December 2017, was the iCIRRUS showcase demonstrator that simultaneously
demonstrated a variety of evolved‐fronthaul lower layer splits and cloud‐assisted Unified
Communication applications.
The demonstrators are KPI validations in themselves as implementation of 5G proof of concept
demonstrators is recognised as a measure of the KPI “European availability of a competitive
industrial offer for 5G services and technologies,” which links to both B3 and S3 business and
societal 5G KPIs respectively [1].
The use cases explored in the demonstrators and the KPIs used to validate their performance are set
out and are related to the overall 5G‐compliant architecture that has been developed by the project.
In this deliverable, the architecture of the showcase demonstrator is illustrated and its component
parts are described, which includes the physical extent of the fibre network that supports the
fronthaul/backhaul.
The performance of the component fronthaul use‐cases and the results obtained for transport over
a 100G aggregated Ethernet link are presented. Respectively, these use‐cases aim to demonstrate: a
low layer split fronthaul for a 60GHz link; CPRI over Ethernet; and femto‐cell backhaul. The
performance achieved by the aggregated link itself is then presented, including synchronisation and
latency.
The results of a separate demonstrator to evaluate the performance of a higher layer split LTE link
over point‐to‐point fibre and PON is presented.
Moreover, the performance of the mobile cloud (Unified Communications) use‐cases, and the
results obtained when these are supported by the LTE link operating over the 100G aggregated
fronthaul are presented.
Finally, the results achieved in all the experiments and their relation to KPIs are summarised, and it is
concluded that they support the assertion that the iCIRRUS concept is valid and has potential for
deployment in 5G networks.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 5 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Indexofterms
3GPP 3rd Generation Partnership Project
5G 5th Generation Mobile Radio Technology
BBU Baseband Unit
BER Bit Error Rate
BS Base Station
BTS Base Station
BW Bandwidth
CAPEX Capital Expense
CFO Carrier Frequency Offset
CO Central Office
CoMP Coordinated MultiPoint
CoS Class of Service
CPRI Common Public Radio Interface
CPRIoEth CPRI over Ethernet
CPU Central Processing Unit
C‐RAN Cloud Radio Access Network
C‐SON Centralised SON
CU Central Unit (abstraction of BBU pool for alternative functional splits)
D2D Device‐to‐Device
D2I Device‐to‐Infrastructure
DAC Digital to Analogue Converter
DBA Dynamic Bandwidth Allocation
DL Downlink
DSCP Differentiated Services Code Point
D‐SON Distributed SON
DSP Digital Signal Processing
DSPLL Digital Signal Phase Locked Loop
DU Digital Unit
eMBB enhanced Mobile Broadband (5G use‐case)
eNB Enhanced Node B (radio base station unit for LTE)
EPC Evolved Packet Core
F1 3GPP NR interface between CU and DU
FAX Facsimile
FDV Frame Delay Variation
FEC Forward Error Correction
FH Fronthaul
FIFO First In, First‐Out
FLR Frame Loss Rate
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 6 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
FMC Fixed Mobile Convergence
FPGA Field Programmable Gate Array
FPU Floating Point Unit
FTTH Fibre to the home
FTTx Fibre to the x
GMC Grand Master Clock
GNSS Global Navigation Satellite System
G‐PON Gigabit Passive Optical Network
GPS Global Positioning System
GST Guaranteed Service Traffic
HARQ Hybrid Automatic Repeat Request
HetNet Heterogeneous Network
HLS Higher Layer Split
H‐SON Hybrid SON
ICT Information and Communication Technology
ID Identifier
IEEE 1588 Ethernet timing protocol
IETF Internet Engineering Task Force
IF Intermediate Frequency
IFDV Inter Frame Delay Variation
IP Internet Protocol
IPDV Inter Packet Delay Variation
iRRH Intelligent Remote Radio Head
IT Information Technology
KPI Key Performance Indicator
LLS Lower Layer Split
LTE Long Term Evolution
MAC Media Access Control
MB Mobile Cloud
MCS Modulation and Coding Scheme
MEC Mobile Edge Computing / Multiple access Edge Computing
MIMO Multiple‐Input Multiple‐Output
MME Mobility Management Entity
MMW Millimetre Wave
NGFI Next‐Generation Fronthaul Interface
NIC Network Interface Card
NID Network Interface Device
NOC Network Operations Centre
NR 5G New Radio
ns Nano second 10‐9
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 7 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
OAI Open Air Interface
OAM Operations, Administration and Management
OC Ordinary Clock (may be master or slave)
ODN Optical Distribution Network
OLT Optical Line Termination
OMC Operations and Maintenance Centre
OPEX Operating Expense
OS Operating System
OSS Operation Support System
OWD One Way Delay
PDCP Packet Data Convergence Protocol
PDG Packet Delivery Gateway
PDV Packet Delay Variation
PGW Packet Gateway
PHY Physical (Layer)
PLL Phase‐Locked Loop
PM Performance Management
PNF Physical Network Function
PoC Proof of Concept
PON Passive Optical Network
PRC Primary Clock
PRE Packet Routing Engine
PtMP Point‐to‐Multi‐Point
PtMP Point‐to‐Multipoint
PtP Point‐To‐Point
PTP Precision Time Protocol
QAM Quadrature Amplitude Modulation
QoE Quality of Experience
QoS Quality of Service
RACH Random Access Channel
RAN Radio Access Network
RAR RACH Response
RAT Radio Access Technology
RAU Radio Aggregation Unit
RCC Radio Cloud Centre
RE Radio Equipment
RF Radio Frequency
RFC Request for Comment (IETF)
RLC Radio Link Control
RN Radio Network
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 8 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
RoE Radio over Ethernet
RRC Radio Resource Control
RRH Remote Radio Head
RRS Radio Remote System
RRU Remote Radio Unit
RTD Round Trip Delay
RTMP Real Time Messaging Protocol
RTT Round Trip Time
RU Radio Unit (abstraction of RRH for alternative functional splits)
RU Remote Unit
RX Receiver
SDN Software Defined Network
SFP Small Form‐factor Pluggable
SI System Information
SLA Service Level Agreement
SM Statistically Multiplexed (traffic)
SMS Short Message Service
SON Self‐Optimising Network
SP Strict Priority
STD Standard Deviation
SyncE Synchronous Ethernet
TCP Transmission Control Protocol
TCO Total Cost of Ownership
TDD Time Division Duplex
ToD Time of Day
ToR Top of Rack
TSN Time‐Sensitive Networking
TX Transmitter
UC Unified Communication
UDP User Datagram Protocol
UE User Equipment
UL Uplink
URLLC Ultra‐reliable and low‐latency communications
vBBU Virtual Baseband Unit
VIP Very Important Person
VLAN Virtual Local Area Network
VNF Virtualised Network Function
vNIC Virtual NIC
WDM Wavelength Division Multiplexing
WRR Weight Round Robin
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 9 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
X2 Interface between eNB
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 10 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Contents
1 Introduction .................................................................................................................................. 12
2 iCIRRUS use‐case and KPI review .................................................................................................. 13
Use cases and societal/business KPIs ................................................................................... 13
Ethernet evolved‐fronthaul use cases .......................................................................... 14
Unified Communications and Mobile Cloud/Clone use cases ...................................... 15
Device‐to‐Device (D2D) use case .................................................................................. 18
Performance KPIs and SLA overview .................................................................................... 19
KPIs for evolved‐fronthaul use‐cases ............................................................................ 19
KPIs for fronthaul plus RAN optimisation use‐case ...................................................... 21
KPIs for mobile cloud use‐case ..................................................................................... 21
KPIs for D2D communication ........................................................................................ 22
Architecture overview ........................................................................................................... 25
3 Verification through test results and analysis .............................................................................. 26
CPRI over Ethernet transport test and verification in live LTE network ............................... 27
Real‐time fronthaul platform using 60GHz wireless ............................................................. 29
Software‐based‐DU radio‐over‐Ethernet .............................................................................. 31
Testbed integration with sync‐enabled aggregator/switch .................................................. 34
Video streaming in C‐RAN with mobile cloud ....................................................................... 38
Mobile task offloading in C‐RAN with mobile cloud ............................................................. 41
Network controlled D2D use‐case ........................................................................................ 44
Ethernet transport lab tests of new higher‐layer functional split fronthaul ........................ 47
4 Description and evaluation of the final Showcase demonstrator ................................................ 49
Description of the Showcase demonstrator ......................................................................... 49
CPRI over Ethernet test and verification in the Showcase ................................................... 52
Real‐time fronthaul platform using 60GHz wireless with UPPER‐PHY LLS verification in the
Showcase .......................................................................................................................................... 55
Reference measurement (fronthaul link only).............................................................. 55
Inclusion of aggregator switches into 60GHz fronthaul link ......................................... 56
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 11 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Inclusion of 3km installed fibre into fronthaul link ....................................................... 58
Software‐based‐DU radio over Ethernet with MAC‐PHY LLS validation in the Showcase .... 60
Femto‐cell fronthaul/backhaul aggregation validation in the Showcase ............................. 65
Aggregated Ethernet‐based fronthaul network validation in the Showcase ....................... 66
Evaluation of the Showcase demonstrator ........................................................................... 67
5 Conclusion ..................................................................................................................................... 70
6 References .................................................................................................................................... 71
7 List of figures ................................................................................................................................. 73
8 List of tables .................................................................................................................................. 75
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 12 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
1 IntroductionThe iCIRRUS project pioneered the proposed use of Ethernet in the fronthaul, and the use of the
intelligent monitoring of this fronthaul, largely enabled by the adoption of Ethernet‐based
technology, to enhance service provision. Of course, in the time since the inception of these ideas,
there has been a large effort globally on 5G, including in the European 5G PPP projects. With these
major developments, it has been necessary to continually place the iCIRRUS effort in context – in
order that it can positively make its own distinct contributions within European and global research
and innovation and standardisation efforts. For example, the iCIRRUS project proposal idea on using
“midhaul”, continued to develop and became part of the efforts in several standards bodies of
defining new Radio Access Network (RAN) functional splits, while intelligent control of these splits
became part of the global moves for 5G of functional virtualisation, orchestration and software‐
defined networking. Similarly, the proposals for using network intelligence for mobile “clone”
operation and offloading processing from mobile devices, and/or offloading traffic from the
network, can be seen in the context of 5G proposals for increased use of Mobile Edge Computing.
These examples demonstrate that iCIRRUS has retained relevance in the rapidly developing 5G
landscape, and it has made significant inputs by focussing on specific challenges (for example,
Ethernet, rather than the full scope of possible fronthaul technologies). Of course, to demonstrate
the significance of these inputs it is necessary to verify performance and, if possible, validate the
fundamental concepts that have been proposed. In order to do this, in this deliverable report, we
examine performance verification in terms of some of the key performance indicators (KPIs) that
have been proposed for the iCIRRUS architecture, in iCIRRUS deliverables D2.1 [2],D2.2 [3], D3.3 [4]
and D3.4 [5] . Further, to put our work in the context of the overall European 5G effort, we relate
these KPIs to those defined by the Euro‐5G project for the 5G PPP [1].
While the primary target has been to verify KPIs, and evaluate these in the context of 5G PPP KPIs to
provide a validation of the iCIRRUS concepts, it has also been important for the iCIRRUS project to
build a final, showcase demonstration. The integration of different, specific technological solutions
that is necessary for such a demonstration also provides a further validation of the architecture,
showing that one technological solution is not a hindrance to another, rather that they are
complementary. Due to verification experiments being available at different times during the
project, and, sometimes unforeseen, difficulties in putting theory and concepts into practice or in
transporting experiments to the final demonstration site in Telekom Slovenije’s labs in Ljubljana, the
showcase demonstration does not attempt to integrate every technological research advance made
in the project; it does, however, integrate both a range of fronthaul technologies over Ethernet, and
cloud/clone services.
In the following Section, the use‐cases explored in iCIRRUS are introduced, the societal and business
KPIs are considered and the performance the KPIs associated with validating their performance are
described, and the interfaces on which measurements are made to determine the KPIs are
illustrated with reference to the iCIRRUS high‐level architecture. The iCIRRUS project’s KPIs are
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 13 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
related to 5G PPP KPIs [1]. In Section 3, the test measurement results obtained to verify the
performance KPIs are described. Section 4 presents an overall evaluation of the showcase
demonstration and the verification results obtained, with a view to validating the iCIRRUS
architecture. Conclusions on the work are presented in Section 5.
2 iCIRRUSuse‐caseandKPIreviewIn this Section, we examine the use cases for the deployment of an iCIRRUS‐type network together
with a comparison with business and societal KPIs and the performance KPIs that have been
targeted by the project and their relation to KPIs defined by 5G PPP. Additionally, the interfaces and
resource clouds on which the quantitative performance KPIs may be measured are considered by
reference to the high‐level iCIRRUS architecture.
Usecasesandsocietal/businessKPIs5G PPP has identified several business and societal KPIs [1]. The first two business KPIs are specific to
the 5G PPP initiative. The only contribution that iCIRRUS can make is indirect through cooperation
with 5G PPP projects (which has been done in several cases, such as joint workshop coordination
and joint contributions to book chapters and other publications). The third business KPI (B3) targets
the market share for EU‐headquartered ICT companies for 5G equipment and services. This is seen
to be related to the third societal KPI (S3) on the “European availability of a competitive industrial
offer for 5G systems and technologies”. The 5G‐Euro project states that there are difficulties in
assessing these KPIs. However, among the assessment measures proposed are (1) the number of
proof of concept demonstrations, (2) contributions to standards, (3) patents, and (4) the costs of
ownership (TCO) and rollout for different coverage targets for operators when using the 5G
technologies developed.
The other societal KPIs include: “Enabling advanced user privacy” (S1); “Reduction of energy
consumption per service…” (S2); “Stimulation of new economically‐viable services of high societal
value…” (S4) and “Establishment and availability of 5G skills development curricula (in partnership
with the EIT)” (S5). In iCIRRUS, the reduction of energy consumption is certainly an important target.
5G‐Euro states that it is a difficult KPI to measure and suggests that it could be done through proxies,
such as spectrum efficiency, specific protocols to switch devices/cells off and energy consumption
reduction in hardware. The other societal KPI with relevance to iCIRRUS is that for the stimulation of
new services.
In our overview of these KPIs for the iCIRRUS project, we examine in turn three areas for the work in
the following sub‐sections.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 14 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Ethernetevolved‐fronthaulusecasesAs has been stated in [2] and [5], the Ethernet evolved‐fronthaul use‐case has several drivers: the
desire to exploit the ubiquity and potential for low‐cost implementation and management of
Ethernet; the need to create a fronthaul solution that is scalable to the throughputs required for the
5G enhanced mobile broadband (eMBB) use‐case; and the desire to exploit statistical multiplexing
for greater transport network efficiency (of use and of energy) . Early work in iCIRRUS and other
projects made it clear that strategies based solely on the transport of time‐domain IQ samples of the
radio baseband waveforms do not scale in an economically efficient manner; however, there still
exist use cases where such transport is required, due to backward‐compatibility with legacy
equipment and to the fact that new functional splits in the RAN do come with potential performance
penalties. The drivers for the new Ethernet evolved‐fronthaul can be summarised as follows:
Reduce fronthaul transport deployment costs
Reduce fronthaul transport management costs
Exploit the ubiquity of Ethernet
Minimise the need for extra cost and complexity in supporting fronthaul requirements
Reduce TCO and rollout costs for different coverage targets for operators
These can be seen very clearly to contribute to the business and societal KPIs (B3, S3) for reducing
cost of ownership and deployment costs. Our work in D2.3 [6] presented an analysis of technologies
to support low cost initial roll‐out of 5G. Factors considered included the expected increase in traffic
demand driven by video services, network sharing, and that the highest bandwidths are not required
to be serviced uniformly across the geography, according to subscriber modelling by Orange,
Telekom Slovenije and PrimeTel. Re‐purposing of existing fronthaul with PtP fibre and PON, and low‐
cost 100G links were addressed and system costs estimated. Also, business cases to estimate the
payback period for deployment of D2D and clone‐based technologies were developed that indicated
payback within 3 years.
The fronthaul work has also contributed to the following standards:
The iCIRRUS evolved fronthaul architecture and concepts have been presented both directly
and indirectly (through discussions during meetings) in several IEEE 1914 working group
meetings.
Many iCIRRUS concepts, and initial implementation have been presented in the
OpenAirInterface software alliance.
ETSI Zero Touch Service and network Management (ZSM) “RAN SON ZSM input iCIRRUS” [7]
Presented iCIRRUS concepts joint fronthaul‐RAN SON. The group requested this material is
used in work items on slicing, developing user stories and use cases and to help align
nomenclature.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 15 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
The following patents have been filed by iCIRRUS partners in relation to the fronthaul work:
“Enhancing Network Topology Information for a Self‐Organising Network,” [8]
“Packet colouring for data stream monitoring,” [9]
“Techniques for providing front‐haul data awareness” [10]
“Method and apparatus for mitigation of packet delay variation,” [11]
Finally, the iCIRRUS fronthaul work has contributed to the business/societal 5G PPP KPIs related to
European proof‐of‐concept demonstrations. While individual partners have focussed on testbeds to
examine performance KPIs, the final showcase demonstration of the project integrating the work of
individual partners in Telekom Slovenije’s lab on Ljubljana is a major proof‐of‐concept for fronthaul
integration. The showcase demonstration integrates transport for two lower‐layer RAN functional
split points (one in upper PHY layer for a wide, 5G‐like bandwidth 60GHz system, the other at the
MAC‐PHY interface in an LTE system), for generic and CPRI IQ waveform transport, and for backhaul.
All are passed through a high‐speed aggregator with timing/synchronisation signals and using time‐
sensitive networking. In addition, an earlier proof‐of‐concept demonstrator was made in Orange’s
laboratory in Lannion, demonstrating a higher‐layer functional split (PDCP‐RLC split in LTE).
The performance KPIs for each of the fronthaul testbeds and the showcase demonstration are
discussed in Section 2.2.
UnifiedCommunicationsandMobileCloud/CloneusecasesLet us first consider what is generally understood by the term Unified Communications (UC):
“Unified communications (UC) is a business term describing the integration of enterprise
communication services such as instant messaging (chat), presence information, voice
(including IP telephony), mobility features (including extension mobility and single number
reach), audio, web & video conferencing, fixed‐mobile convergence (FMC), desktop sharing,
data sharing (including web connected electronic interactive whiteboards), call control and
speech recognition with non‐real‐time communication services such as unified messaging
(integrated voicemail, e‐mail, SMS and fax). UC is not necessarily a single product, but a set
of products that provides a consistent unified user interface and user experience across
multiple devices and media types” [12].
Consequently, from the business point of view it is appropriate that its performance measures and
goals are directed towards improving business targets such as competitiveness, excellence in the
attention to clients, and increase in network efficiency and effectiveness, while minimizing the
consumption of resources.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 16 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
In the framework of ICT, the UC combines all the communications possibilities onto one platform,
instead of having them scattered and separate. UC is not a concrete product: Unified
Communications is a basic and strategic solution for any company. Some studies have revealed that
more than 60% of the employees of a company manage different communications devices. Having
to manage different media on different platforms considerably reduces the effectiveness of
communications within a company.
Unified Communications unifies each means of communication in a company (documents exchange,
calls, message, emails, etc.). These are grouped in an intelligent manner, increasing user
productivity, and offering better solutions with better quality that improves user’ experience. In
addition, the UC platform offers ubiquitous accessibility, so that UC systems may be provided for any
business model, regardless of its size, resources and location.
Concerning the iCIRRUS objective, this use case can be considered as a proof of concept that adds
value to 5G networking through communications in the cloud. The mobile cloud is also an added
value to the UC because it offers elasticity and agility, with companies able to scale up and down the
functionalities on offer, as needed. The potential advantages offered by UC with the mobile cloud in
comparison with traditional hosted services include:
Reduce IT infrastructure cost
Increase business elasticity and scalability: provisioning the resources as needed
Standards and flexibility
Multi‐tenancy
Disaster recovery
In this sense, the UC system provided by Wellness Telecom benefits from the iCIRRUS project with
the indirect benefits of a 5G communication system. In a voice and real‐time application, the
advantages presented by iCIRRUS are experienced by users in the QoE. To see the impact of this
pilot, different UC platforms are compared and the final value proposition of this use case is
presented.
Multiple platforms are available in the market, with a selection of these being summarised in Table
2‐1. As the feature terminology is not standardised, the Table can only provide a rough comparison
of feature comparability.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 17 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
presence
chat
fax
voicemail
IM
Video
bridge
Meetings
Documen
t
man
agement
IP telephony
fixed m
obile
integration
Platform
sm
artphone
Platform
PC
Cloud support of
application
Elastix1 x x x x x X x
Cisco2 x x x x X X x
Mitel3 x x x X x x
Nortel4 x x x X x x x
Innova‐phone5 x x x x X x x
Telefonica6 x x x X x x X x
WT iCIRRUS x x x X x X x X
Table 2‐1 Comparison of Unified Communication offerings
Within iCIRRUS two mobile cloud/clone use cases were integrated over the fronthaul of the
showcase demonstrator: the “Video streaming in C‐RAN with mobile cloud,” and “Mobile task
offloading in C‐RAN with mobile cloud.” The associated performance KPIs are described in Section
2.2. The integration of these service‐type technologies with the lower‐layer networking technologies
of the fronthaul is a further indication of the value of the iCIRRUS showcase demonstration. These
are illustrated in Table 2‐1 in comparison with the other UC offerings, illustrating how use of cloud
support of applications by iCIRRUS is a differentiator that may help deliver 5G KPIs.
UC is indispensable in the business world, improving reachability, increasing efficiency and,
accelerating business process. However, such a UC service tends to consume a great quantity of the
resources available in a device, eg processing large amounts of video and audio data. The iCIRRUS
system directly benefits Unified Communications using the Mobile Cloud with the potential to
improve the QoS in two ways:
Improving the provisioning of voice and video quality in a 5G network by reducing latency
and increasing bandwidth was proved in the offloading task in D5.3 [13].
The consumption of resources in the devices is reduced because most of the computing
workload is carried out in the cloud (as explained later when examining the performance
KPIs).
1 https://www.elastix.org/.December 2017 2 https://www.cisco.com/ December 2017 3 http://www.mitel.com/ December 2017 4 http://www.minitel.ca/Files/CMFiles/46UnifiiedCommunications.pdf December 2017 5 http://www.innovaphone.com/en/produkte-unified-communications.html December 2017 6 http://www.movistar.es/grandes-empresas/soluciones/fichas/comunicaciones-unificadas/ December 2017
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 18 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
The Wellness Telecom UC service offering can certainly be an innovative, and competitive 5G service
(S4): WT expect benefits from iCIRRUS with this UC pilot of 70.000 € by 2020. There are aspects of
what can be provided, for example, disaster‐recovery, that undoubtedly have high societal value.
The media mixer implemented and demonstrated in D5.3 [2] processed all the video and image data,
resulting in improved battery life of the devices. Thus, there is a contribution to energy consumption
reduction and the 2nd societal KPI “reduction of energy consumption per service” [1] (S2). The
application demonstration proved the reliability of the 5G communications network connected with
the Mobile Cloud, one of the aims of the iCIRRUS project, and contributing also to the service
viability societal KPI (S4).
Contributions to standards and patents, were not considered appropriate to the chosen business
model which is to deliver the product via 3rd party software.
Device‐to‐Device(D2D)usecaseDevice‐to‐device (D2D) communications will serve to meet business/societal KPIs in the following
ways:
Improved spectrum and energy use for operators (S2)
Improved energy efficiency for user devices (S2)
Stimulation of new service offerings (S4)
Extension of coverage to facilitate reduced TCO (S3) and attainment of reliability and
accessibility KPIs
The work in iCIRRUS has led to filing of a patent application:
“High resolution angle of arrival estimation and dynamic beam nulling”, [14]
which discloses methods for establishing very high throughput communication links between
devices using novel millimetre wave signal angle of arrival estimation. The application also provides
practical procedures to minimize interference between devices using transmitter side beam
optimization techniques. The patent follows work done in the IEEE 802.11ay standards group on
millimetre wave channel characterisation. IDCC envisage this work as building on their previous D2D
communication platform operating at microwave/RF.
In iCIRRUS, the aim was that the above benefits would be further enhanced by examining
possibilities for using the remote radio units in D2D functions, such as enhancing user device
localisation and, with higher layer functional splits, moving decision‐making for D2D offloading from
the mobile network closer to the users, allowing it to be more responsive. Performance KPIs for D2D
are examined in Section 2.2.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 19 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
PerformanceKPIsandSLAoverviewThe following Sections outline the performance KPIs for the iCIRRUS architecture, taking the
evolved‐fronthaul use cases, the mobile cloud/clone and D2D use cases in turn. KPIs for the joint
optimisation of the fronthaul and mobile RAN are also considered.
From Euro‐5G/5G PPP [1], 4 over‐arching performance KPIs were identified.
The first was in “providing 1000 times higher wireless area capacity and more varied service
capabilities”. Euro‐5G translated this into two distinct KPIs:
P1a – absorbing Tb/s traffic in a smart office (equivalent to 10Tb/s/km2 wireless area
capacity)
P1b – reaching a peak user data rate of between 1 and 10Gb/s for specific scenarios/use
cases
The second KPI was on reducing the average service creation time to 90 minutes. Euro‐5G proposed
to monitor the following for this:
P2a – level of automation of service processes
P2b – existence/range of autonomic network management functions
P2c – use of Open Source, SDKs etc., to facilitate deployments
The third performance KPI was on facilitating very dense deployments for large numbers of users
(people and devices). These were translated to:
P3a – handling up to 1 million devices per km2 in certain scenarios/use cases
P3b – facilitating deployment and operation of a large number of small cells (tens of sites
per km2)
The fourth performance KPI was in providing a secure, reliable and dependable internet. This was
translated to:
P4a – providing a latency of between 1 and 10ms (1 ms being for the air interface link)
P4b – reaching reliability >99% or more (in terms of packets delivered within latency budget)
P4c – reach an availability >99% (in terms of users being covered by a 5G service)
KPIsforevolved‐fronthauluse‐casesOne of the primary transport related use‐cases that iCIRRUS has investigated is that of the evolved‐
fronthaul for a RAN with different functional splits. Table 2‐2, adapted from D3.4 [5] with added
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 20 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
columns to include the backhaul case and the 5G KPIs, is a compilation of the fronthaul KPIs for the
different use‐cases that are, respectively: transport of legacy CPRI; two examples7 of transport with
a low‐layer‐split (LLS) with and without CoMP support; and an example of a higher‐layer‐split (HLS).
With the concurrent development of standards and quasi‐standards during the unfolding of the
iCIRRUS project, these respective use‐cases may be related with reasonable accuracy to
(1) the LLS that is being addressed by the eCPRI industry “standard,” that is designed to target
requirements for the LLS.
(2) the LLS split options that are being discussed within 3GPP
(3) the HLS that is being standardised by the 3GPP as the f18 interface between the centralised
unit (CU) and distributed unit (DU).
Different types of traffic supported by the evolved‐fronthaul
Fronthaul Requirements
Legacy traffic (CPRI
Upper PHY split in DL and UL (no CoMP)
Upper PHY split in DL lower PHY split in UL (CoMP)1
PDCP‐RLC split Backhaul (unified transport)
5G KPIs
User Data rate 100‐400Gbps
(P1) Area coverage
(P1a) 10TB/s/km2
(P1b) Per user 1‐10GB/s
FH data rate
User Latency (P4a) 1‐10ms 9
Max latency (round trip delay)
150µs (CoMP) 440µs (no CoMP)
440µs 150µs 60ms 60ms
Min frequency accuracy
+/‐ 2ppb (per hop)
+/‐ 2ppb (per hop)
+/‐ 2ppb (per hop)
+/‐ 2ppb (per hop)
Min phase and timing accuracy
+/‐ 10ns (MIMO & TX diversity)
+/‐ 30ns +/‐ 30ns
TBD (expected approx. +/‐
1.36s (LTE TDD)
Max latency imbalance
+/‐ 16ns +/‐ 163ns +/‐ 163ns Not yet defined
Max BER 10‐12 10‐6
10‐3 non‐priority data 10‐5
priority data
[1] The MAC‐PHY split has similar KPI requirements so they are not separately presented here
Table 2‐2 Compilation of Fronthaul KPI
7 One option is MAC-PHY and the other is UPPER-PHY 8 A parallel interface is being standardised for 4.5G called the v1 interface 9 1ms being for the air-interface link. User plane 1-way delay <4ms eMBB, <1ms URLLC
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 21 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Table 2‐2 defines the KPIs that are applicable for the Ethernet fronthaul scenarios in Section 2.1.1,
including the transport of CPRI, HLS and LLS, and femto‐cell backhaul and by extension to any
aggregation in a 100Gb/s+ fronthaul scenario that would still have to meet these requirements. In
terms of higher level performance KPIs, fronthaul KPIs impact particularly on latency considerations.
The Euro‐5G document gives examples of processing delays, synchronisation and retransmissions as
steps that need to be quantified within User Plane latency as an illustration of performance
assessments that would have to be carried out.
KPIsforfronthaulplusRANoptimisationuse‐caseAs an extension of the evolved‐fronthaul use‐case, iCIRRUS has also addressed the use‐case for joint
optimisation of the fronthaul and RAN. This required an increase in the number of KPIs that should
be monitored; however, these are RAN performance indicators D3.3 [4], rather than minimum
acceptable criteria for operation, as in Section 2.2.1.
KPI Definition 5G KPI target
Accessibility Success rate of access attempts to the system
(P4a) >99%
Retainability 1 ‐ dropped call rate (P4b) >99%, URLLC 99.999%10
Throughput Median and 95th percentile throughput
(P1b) 1‐10GB/s
Application performance Per application, eg mean opinion score for voice
Fronthaul redundancy Measure of available alternative paths
Figure 2‐1 KPIs for fronthaul and RAN joint optimisation
These system‐level KPI goals are not easy to quantify directly without a large‐scale system simulation
or wide‐area deployment. However, the “retainability” KPI, which is a common performance KPI in
today’s mobile networks that stems from a circuit‐switched view of the world, can be mapped to
“reliability.” As, “reliability,” is defined as successful delivery percentage of packets within a latency
constraint, it allows us to relate the KPI, to some extent, to achievement of the enhanced fronthaul
latency KPIs. Additionally, it is useful to state these KPIs here, to provide context for the ultimate
goals of an iCIRRUS derived system deployed in a commercial network.
KPIsformobileclouduse‐caseThe mobile cloud use‐case in the iCIRRUS project is explored with two proof‐of‐concept (PoC)
applications as examples of possible components in an enhanced UC offering: video streaming
supported by the mobile cloud, and mobile application offloading supported by the mobile cloud.
10 Packets delivered successfully within the required latency
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 22 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
These use‐cases can be mapped to the business KPI B3 to the extent that they have potential to
impact TCO and there is a direct mapping to societal 5G KPIs related to S4 “stimulation of new
economically‐viable services…” and S2 “the reduction of energy consumption…” The mobile cloud
defined for these services has linkage to the 5G performance KPI P2 “reducing the average service
creation time cycle”
KPIsforD2DcommunicationD2D communication in iCIRRUS allows two devices in proximity to set up a direct D2D transmission
link underlaying the iCIRRUS network infrastructure, allowing offloading of traffic from this
infrastructure. However, to guarantee the quality of service (QoS) of D2D communications,
significant assistance is required from the C‐RAN architecture, including D2D link set up and resource
management. According to the heterogeneity of D2D communication, in which D2D communication
is a potential alternative to cellular communication, strict numeric KPIs may vary for different use
cases. However, several KPIs could be used to demonstrate the requirements and benefits of D2D
communication in the iCIRRUS architecture, as shown in Table 2‐3 and Table 2‐4. In previous works,
reported in D5.3, the tests for D2D set up along with the real‐time video transmission via D2D
communication were discussed [13].
KPI for D2D link set up Definition 5G KPI target
Set up latency The time consumption of the signalling overheads between D2D pair and BBU pool via fronthaul.
(P4a) CP latency < 10ms UP latency 1‐10ms
Discovery range The maximum transmit‐power of pilot signal, which determines the maximum D2D set up range.
(P3a) 10,000‐1,000,000 per km2 devices
Discovery efficiency The ratio of the number of successful D2D set‐ups to the number of D2D requests.
(P3a) 10,000‐1,000,000 per km2 devices, (P4b) >99%
Table 2‐3 KPIs for D2D link set up
D2D link set‐up in iCIRRUS is related to 5G KPIs for control plane. The detailed explanation of KPIs for
D2D link set‐up are shown below:
1. Set‐up latency: From the D2D set‐up framework mentioned in D2.2 [15], several handshakes
are required between the D2D pairs and BBU pool. For each handshake, the delay between
UE and BBU pool significantly impacts the D2D link set‐up latency; hence the number of
handshakes should be limited to guarantee the QoS of D2D communication. In iCIRRUS, a
location based D2D discovery and link set‐up strategy has been proposed. Compared to the
standard discovery strategy which involves the communications with the BBU pool at least
three times and includes 7 or more steps, our proposed strategy only needs to communicate
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 23 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
with the BBU pool once and has only 4 steps for D2D link set‐up. It makes our contribution
related to the 5G KPI P4a on “latency.”
2. Discovery range: The maximum transmit power of the pilot signal of the user device
determines the maximum D2D set‐up range. It could be adjusted dynamically according to
the specific service requirements of D2D communication, eg data rate, latency, etc.
3. Discovery efficiency: A key performance metric to evaluate the discovery strategy is the
successful set‐up ratio, which leads our contributions to the 5G KPI P4b “reliability.”
Furthermore, for real‐time D2D communication, the D2D discovery protocol proposed in
iCIRRUS is an asynchronous one. As the locations of UEs can be proactively estimated at
their access RRSs, the resource blocks (RBs) and power allocation for pilot transmission,
which is used for D2D link set‐up, could be optimized to increase the potential matching of
D2D pairs. Meanwhile, the spectral and energy efficiency of the discovery process can be
improved.
KPI for D2D communication Definition 5G KPI target
UE data rate Data rate received at the
receiver of D2D pair in bits per second.
(P1b) reaching a peak user data rate of between 1 and
10Gb/s
Transmission efficiency The packet delay and loss rate
between two devices.
Energy efficiency Power consumption at each
device in bits per joule.
Resource management latency The time consumption of one signalling packet for D2D resource management.
(P4a) CP latency < 10ms UP latency 1‐10ms
Table 2‐4 KPIs for D2D communication
D2D communication in iCIRRUS is related to 5G KPIs for both data plane and control plane. The
detailed explanation of KPIs for D2D communication in iCIRRUS are shown below:
1. UE data rate: Due to the short transmit distance of D2D communication, the received data
rate can be greatly enhanced. Therefore, D2D communication can be used for large volume
content delivery, such as high‐quality video transmission. The proposed semi‐distributed
resource management in iCIRRUS could improve the D2D data rate by several times through
the introduction of frequency reuse in a BBU’s coverage area, which can contribute to the
5G KPI P1b.
2. Transmission efficiency: The packet delay and loss rate of D2D communication are much
lower than in cellular communication, which makes D2D a vital network component to
enable some latency‐critical applications, eg online real‐time gaming, virtual reality, mobile
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 24 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
edge computing, etc, that may contribute to the 5G societal KPI to simulate new services S4
as well as latency with P4a.
3. Energy efficiency: With direct data transmission between two devices, the transmit power at
transmitter could be greatly reduced to achieve desired QoS in various use cases, which
contributes to societal 5G KPI S2.
4. Latency for resource management: It is essential to explore network assisted resource
management for the D2D communications to achieve optimal network performance, eg
throughput, outage probability, etc. Therefore, the duration of the signalling between D2D
pairs and BBU pool might exceed the maximum delay requirement of some real‐time
services. In iCIRRUS, it is considered that some of the MAC layer function blocks for D2D
resource allocation is shifted from the BBU pool to RRHs through the functional splitting in
the fronthaul, thus reducing significantly the resource management latency, which
contributes to the 5G KPI P4a.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 25 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
ArchitectureoverviewThis Section maps the quantitative KPIs introduced in the previous Sections to the interfaces and
resource clouds of the iCIRRUS conceptual architecture, illustrated in Figure 2‐2, at which these
quantitative KPIs values may be measured. The actual means of determining the KPIs is described in
Section 2.1.
Figure 2‐2 iCIRRUS architecture overview
Considering the evolved‐fronthaul use‐cases outlined in Section 2.1.1,
IQ Transport Fronthaul, is shown in Figure 2‐2, emanating from the BBU pool, conveyed via
one or more Ethernet switches and delivered to the RRHs towards the top right of the
Figure. The respective end‐points for measuring the KPIs set out in Table 2‐2 are the BBU
pool and the concerned RRH.
NGFI‐type fronthaul, is shown in Figure 2‐2, emanating from the RCC pool, conveyed via one
or more Ethernet switches and delivered to the RRSs (which may themselves be split
between RAU and RRU) towards the bottom left of the Figure, the respective end‐points for
measuring the KPIs set out in Table 2‐2 are the RRC pool and the concerned RRS/RRU. The
Figure does not show the details of the functional split in the radio so LLS and HLS use‐cases
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 26 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
are not distinguished. The respective end‐points for measuring the KPIs set out in Table 2‐2
are the RCC pool and the concerned RRS/RRU.
Backhaul aggregation with fronthaul, is shown in Figure 2‐2, with backhaul emanating from
the RCC pool and conveyed via one or more Ethernet switches through the “backhaul to
core”. The end‐points for measuring KPIs are the RCC pool and the core network.
Aggregation of “midhaul” between the micro BTS at the bottom of the Figure and the RCC
pool with fronthaul and backhaul is also shown. The respective end‐points for measuring the
KPIs set out in Table 2‐2 are the Micro‐BTS and the RCC pool.
The aggregated Ethernet fronthaul is represented by the vertical trunks that interconnect
the Ethernet switches and the Ethernet links connecting Ethernet switches to the RCC pool
in Figure 2‐2. The performance of the aggregated fronthaul is verified by the simultaneous
meeting of the KPIs of the component use‐cases.
Considering the Unified Communication example use‐cases examples outlined in Section 2.1.2,
The mapping of the UC use‐case components to the elements of the iCIRRUS architecture in
Figure 2‐2, is the same for both “Video streaming in C‐RAN with mobile cloud,” and “Mobile
task offloading in C‐RAN with mobile cloud.” The application component is hosted in the
mobile cloud, that is in MB1 or MB2 on the cloud side and in the mobile device itself on the
other. The real‐time telemetry framework which functions as a virtual infrastructure
manager (VIM) for collecting the KPIs in these component in the Mobile Cloud is described
respectively in Sections 4.6 and framework collecting and storing phone related KPIs is
described in Section 3.5.
Considering the network controlled Device‐to‐Device example use‐case examples outlined in Section
2.1.3,
The network controlled D2D use‐case components are mapped to the elements of the
iCIRRUS architecture in Figure 2‐2 as follows: the network controlled element maps to the
MAC that may be in the RCC pool. Alternatively, the MAC element dealing with D2D may be
pushed nearer to the network edge at the RAU. The D2D element is mapped to the two or
more mobiles located at the right of the Figure.
3 VerificationthroughtestresultsandanalysisSection 3 summarises the results of the various testbed experiments conducted throughout the
iCIRRUS project and compares them with the iCIRRUS design target KPIs and also validates them in
the broader context set by the 5G business, societal and performance KPIs set out in [1].
Section 4 addresses the Showcase demonstration where a subset of the individual testbeds were
integrated and simultaneously supported on a 100G Ethernet fronthaul link. However, in cases
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 27 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
where the most complete validation of the individual testbeds was conducted at the same time and
location as the Showcase, those Showcase demonstrator results are reported in Section 3 together
with the individual testbed results. Section 4 then refers back to these results in the broader context
of the Showcase.
CPRIoverEthernettransporttestandverificationinliveLTEnetwork
The starting point of the CPRI to Ethernet test was a direct 10G Ethernet connection established
between 2 FPU Boards, see Figure 3‐1. The VIAVI MTS 5800 CPRI Tester acted as both master and
slave device within the CPRI link, using PRBS31 pattern for reasonable transmission tests. The
transportation of frequency information from master to slave was achieved by a synchronous
Ethernet connection (SyncE). With this, frequency recovering was enabled on the BBU side (recovery
of CPRI Master clock) as well as on the RRH side (recovery of Ethernet clock). The BBU side Ethernet
core was driven by a DSPLL clock derived from the recovered CPRI clock. In a comparable manner,
the RRH side’s CPRI clock was generated from the recovered Ethernet clock. As a result, both BBU
and RRH CPRI core clocks operated effectively at the same frequency, and the difference between
the clock frequency on the master side and on slave side couldn’t be measured. The required
minimum frequency accuracy according to Table 2‐2 was met (± 19.66Hz @9.83Gbit / ± 4.92Hz
@2.45Gbit).
iCirrus FPGA Board ‚BBU Side‘
VIAVI MTS 5800CPRI Tester
CPRI Master
Port
CPRI / Ethernet BridgeClock Unit(DSPLL)
RecoveredCLK CPRI
REF CLK
Eth REF CLK
CPRI Core
10G Eth Core
CPRI SlavePort
BERMeasurement
iCirrus FPGA Board ‚RRH Side‘
CPRI / Ethernet Bridge
Clock Unit(DSPLL)
Recovered CLK
CPRIREF CLK
Eth REF CLK
CPRI Core
10G Eth Core
CPRI CPRI
SYNC Ethernet
Figure 3‐1 Point‐to‐point with SyncE
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 28 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Fronthaul KPI Requirement Measured
performance Comment
User Data rate 100‐400Gbps ‐
FH Data Rate
User Latency
Max latency (round trip
delay)
150µs (CoMP) 440µs (no CoMP)
35µs direct link
Exceeds most demanding requirement
Min frequency accuracy
+/‐ 2ppb (per hop)
Better than (± 4.92Hz
@2.45Gbit)
Actual error was less than measurement capability
Min phase and timing accuracy
+/‐ 10ns (MIMO & TX diversity)
Max latency imbalance
+/‐ 16ns Not measured
Max BER Below measurement sensitivity
Table 3‐1 Legacy 2.45Gbit CPRI traffic over Ethernet, requirement v performance
Fronthaul KPI Requirement Measured
performance Comment
User Data Rate
FH Data rate 100‐400Gbps ‐
User Latency
Max latency (round trip delay)
150µs (CoMP) 440µs (no CoMP)
10µs direct link
Buffer depth set by more latency consuming 2.45Gbit
Min frequency accuracy
+/‐ 2ppb (per hop)
Better than (± 19.66Hz
@9.83Gbit)
Actual error was less than measurement capability
Min phase and timing accuracy
+/‐ 10ns (MIMO & TX diversity)
Max latency imbalance
+/‐ 16ns ‐ Not measured
Max BER ‐ Below measurement sensitivity
Table 3‐2 Legacy 9.83Gbit CPRI traffic over Ethernet, requirement v performance
A comparison with the target KPIs is provided in Table 3‐1 and Table 3‐2. The results indicate that
the CPRI over Ethernet can exceed most of the KPI requirements that were identified. The peak
timing accuracy performance as measured on the aggregated link indicates that performance may
not be sufficient for the most severe MIMO and TX diversity requirements.
The benefit of the presented test results is detailed knowledge of the CPRI over Ethernet functioning
as expected, the upcoming mobile generation of mobile fronthaul is expected be based on Ethernet
transport technology. In the network upgrading phase, the interconnection possibility of different
technologies (backward compatibility) is very important to operators. Consequently, demonstration
of CPRI over Ethernet can be related to the business and societal KPI B3/S3 related to cost of
ownership and rollout of 5G with its support of backward compatibility.
With the knowledge obtained of the CPRI/ETH conversion performance and its technical demands, it
is possible to consider it in all planned network maintenance upgrades (to be prepared in advance).
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 29 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
In this way, similar converters could be used for 2G, 3G and 4G remnants on the 5G fronthaul
segment.
Real‐timefronthaulplatformusing60GHzwirelessThis Section summarises the test results of the 60GHz fronthaul system and its integration with the
time sensitive Ethernet aggregated fronthaul link in the ADVA lab that was reported in D5.3 [13] and
compares them with the iCIRRUS design target KPIs and also validates them in the broader context
set by the 5G business, societal and performance set out in [1]. The analysis of the results in the
context of the Showcase demo is presented in Section 4.3.
Figure 3‐2: Test setup for Integration of DU, RU, UE and 60GHz
Figure 3‐2 shows the test set‐up used and Figure 3‐3 summarises the results over the link with
different and varying frame sizes. The lowest latency can be observed for the reference
measurement without the aggregator switch, but with the VIAVI Ethernet protocol analyser (yellow
curve) in all cases. The curves for the test series with the aggregator switch with or without PTP
traffic are almost indistinguishable (violet and green) and show a slight increase of the latency with
respect to the reference measurement. The red curve, with a slightly worse latency, shows the
results of the using a commercial switch instead of the aggregator, however, as the DU and RU
firmware was different, the difference is likely not significant. The highest latency is observed for the
setup with the UE device included.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 30 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Figure 3‐3: comparison of all setups for different frame lengths: 86bytes (a), 512bytes (b), 1024bytes (c), 1518bytes (d), random size 86…1518bytes (e)
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 31 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Fronthaul KPI
Requirement Upper PHY split in DL and UL (no
CoMP)
Requirement Upper PHY split in DL lower PHY
split in UL (CoMP)
Performance 5G KPI target
User Data rate 100‐400Gbps 100‐400Gbps 2.2Gbit/s (P1b) 1‐10Gbit/s
FH data rate
User Latency (P4a) 1‐10ms on the air
interface
Max latency (round trip
delay) 440µs 150µs
250µs @ 1Gbit/s 1200µs @ 2.5Gbit/s
Min frequency accuracy
+/‐ 2ppb (per hop)
+/‐ 2ppb (per hop)
Min phase and timing accuracy
+/‐ 30ns +/‐ 30ns
Max latency imbalance
+/‐ 163ns +/‐ 163ns
Max BER 10‐12
Table 3‐3 lower layer split at the UPPER PHY, requirement v performance
The results are compared with the iCIRRUS KPI targets and the 5G KPIs in Table 3‐3, and the
performance meets the 5G performance KPI for throughput and air‐interface latency depending on
the scenario.
Furthermore, the testbed can also be seen as contributing to the business goal B3 considering the
number of PoC as a proxy for the 5G readiness of the European telecommunications sector, to the
societal goal S3 considering the assessment of the maturity and performance of the different
technical components and the potential impact to TCO, and to the goal S4 considering the potential
to stimulate new high‐value services given the availability of significant allocated bandwidth in the
millimetre wave band.
Software‐based‐DUradio‐over‐EthernetThis Section summarises the test results of the software‐based‐DU radio over Ethernet system that
was reported in D5.3 [13] and compares them with the iCIRRUS design target KPIs and validates
them in the broader context set by the 5G business, societal and performance set out in [1]. The
analysis of the results in the context of the Showcase demo is presented in Section 4.4.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 32 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Figure 3‐4 Indicative setup for software‐based‐DU radio over Ethernet
Figure 3‐4 is indicative of the set‐up used to generate the results summarised here11. Two sets of
results were generated, the first comprising non‐LTE results and the second based on an LTE system.
The non‐LTE results more fully test the capability of the fronthaul as arbitrary data is sent at user‐
defined sampling rates to and from the RRH. All tests were successful in utilising the available
fronthaul bandwidth without the occurrence of buffer underflows or overflows in the DU. For the
second set of results the SW eNodeB and EPC were both active.
11 Several different set-ups were used and this is exemplary as it includes all elements of interest
DU
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 33 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Fronthaul KPI
Requirement Upper PHY split in DL and UL (no
CoMP)
Requirement Upper PHY split in DL lower PHY
split in UL (CoMP)
Performance 5G KPI target
User Data rate 100‐400Gbps 100‐400Gbps 0.05Gbit/s (Non‐LTE) 1
0.02Gbit/s (LTE) 2 (P1b) 1‐10Gbit/s
FH Data rate N/A N/A 0.8Gbit/s (Non‐LTE) 3
0.4Gbit/s (LTE) 4
User latency N/A N/A N/A (P4a) 1‐10ms on the air
interface
Max latency (round trip
delay) 440µs 150µs
40µs (Non‐LTE) 5 266µs (LTE) 5
N/A
Min frequency accuracy
+/‐ 2ppb (per hop)
+/‐ 2ppb (per hop)
N/A N/A
Min phase and timing accuracy
+/‐ 30ns +/‐ 30ns N/A N/A
Max latency imbalance
+/‐ 163ns +/‐ 163ns N/A N/A
Max BER6 10‐12 10‐12 N/A 1. The effective use rate is achieved by scaling the LTE bandwidth by the relative fronthaul rates in D5.3 Table 2‐4 and Table 2‐5 =
800/370 * 20Mbit/s for the X310 radio, D5.3 page 34. 2. This is the user rate from Table 2‐5 for the X310 radio 3. This is the fronthaul data rate in D5.3 table 2‐4 for the X310 radio 4. This is the fronthaul data rate in D5.3 table 2‐5 for the X310 radio 5. The fronthaul latency is a function of the sample width and sample rate, these values are based on the value in D5.3 Figure 24
scaled to the used sample width and sample rate, that is, 173 * 5.76/25 and 173 * 5.76/3.75 6. Performance value based on specification of Ethernet equipment (i.e. not measured)
Table 3‐4 lower layer split at the MAC/PHY, requirement v performance
These tests help to demonstrate different iCIRRUS benefits and innovations, such as the intelligent
processing performed by the intelligence unit to achieve a self‐optimizing network, and data rate
reduction through Functional split(ting) and statistical multiplexing gains, which support the business
goal B3 considering the number PoC, and S3 considering quantified assessment of maturity and
performance of technological components.
In addition, the use of Ethernet in the fronthaul is validated, which is a potentially low‐cost
commodity technology with standardized carrier Class OAM and connection verification facilities. It
can aid the convergence of fronthaul and legacy backhaul (x‐haul), and the insertion of background
traffic was used to aid the evaluation of these convergence scenarios, which supports goal S3
considering TCO.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 34 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Testbedintegrationwithsync‐enabledaggregator/switchBoth local and remote 100G aggregators employed the Guaranteed Service Transport (GST) priority
class (developed by Transpacket12) to establish the PTP link. More details on the GST principle and
implementation can be found in D5.3 [13].
In the integration test, the PTP configuration was the same as that applied in D5.3, and the 1G client
port at the CO was used to forward and broadcast the PTP stream from the GM clock to all the
remote client ports. The PTP analysis at the remote 1G and 10G client ports was performed by the
VIAVI MTS‐5800 tester. Table 2‐1 summarises the PTP link stats for both ports. Note that the PTP
stream egressing from the remote 1G client port was eventually fed into the slave clock (OSA‐5410).
Remote port Master to Slave
delay (µs) Slave to Master
delay (µs) Two‐way
time error (µs)
1G 187.631 188.081 0.225
10G 115.350 115.826 0.238 Table 3‐5 PTP link stats at the remote node
The sync PDV and delay request PDV at the two remote client ports are depicted in Figure 2‐7,
respectively.
1G Port
12 http://www.transpacket.com/
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 35 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
10G port
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 36 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Figure 3‐7 PDV analyses at the remote 1G and 10G ports
Compared to the results in D5.3 [13], apart from the additional propagation delay induced by the
field fibre ring, no other impact was observed in the field integration.
Figure 3‐8 depicts the calculated maximum time interval error (MTIE) and time deviation (TDEV) by
applying a 0.1 Hz first‐order low‐pass filter per ITU‐T G.8271.1 to the measured time error samples.
It can be seen that with the observation interval of wider than 2000s, the MTIE persists at a constant
value of around 110 ns corresponding to a frequency accuracy of less than 55ppt, proving that the
recovered PTP clock at the remote 10G client port was synchronised with the grand master clock
without any frequency offset. Given a shorter observation interval of 10s, the derived frequency
accuracy is then less than 3ppb, however, this value is considered as a pessimistic estimation.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 37 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Figure 3‐8 MTIE/TDEV analysis of PTP time error at remote 10G client port
Table 3‐6 summarises the measured synchronisation performance between two aggregators with
respect to the Fronthaul KPI.
Fronthaul Requirements
Legacy traffic (CPRI)
Upper PHY split in DL and UL (no CoMP)
Upper PHY split in DL lower PHY split in UL (CoMP)
PDCP‐RLC split
Measured performanc
e Comments
Min frequency accuracy
+/‐ 2ppb (per hop)
+/‐ 2ppb (per hop)
+/‐ 2ppb (per hop)
+/‐ 2ppb (per hop)
< 3ppb (2 nodes)
Min phase and timing accuracy
+/‐ 10ns (MIMO & TX diversity)
+/‐ 30ns +/‐ 30ns
Not yet defined (expected approx.
+/‐ 1.36s (LTE TDD)
Average 100ns
Better accuracies require additional means such as active delay measurement and compensation. Note that the GPS reference clock for the probe tester has a maximum time error of +/‐ 50ns.
Max latency imbalance
+/‐ 16ns +/‐ 163ns +/‐ 163ns Not yet defined
Average 200ns
Table 3‐6 Fronthaul synchronisation requirement v performance
Therefore, the GST‐enabled 100G aggregators can broadcast the PTP stream without inducing
additional PDV, which allows the recovered PTP clock at the remote node to be precisely
synchronised with the grand master clock (eg the remote Ethernet‐to‐CPRI clock is recovered from
the PTP clock).
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 38 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
VideostreaminginC‐RANwithmobilecloudIn the video streaming application use‐case, a user of a mobile phone can connect to a mobile clone
that acts as a proxy for access to the video stream in the Internet, with the mobile clone providing a
video transcoding service. The components used in this use‐case are listed in Table 3‐5.
Component Description Implementation
Video server Stream video using RTMP protocol to a local port.
ffmpeg / Avconv
Video proxy Pull a live video stream from remote host, transcode the video stream, then publish the video stream to a local port.
Nginx Nginx RTMP module ffmpeg
Video client Consume the video stream. VLC player
KPI collector Collect the measurements of this use case, such as CPU, network interface traffic.
Snap13 framework Snap plugin collector‐psutil Snap plugin publisher‐influxdb
KPI storage Store the measurements of this use case. InfluxDB
KPI monitoring Publish the measurements of this use case for monitoring.
Grafana
Table 3‐5 Component of the video streaming use‐case
The sequence of steps performed in this use‐case is as follows:
1. The video server (eg ffmpeg) produced an RTMP video stream, which was live streamed and
accessible through a local port, as illustrated in Table 3‐5
2. The mobile phone of a user requested the video stream by sending a request to the mobile
clone
3. The mobile clone requested the video stream from the video server and transcodes the
video stream
4. The mobile phone received the transcoded video stream from the mobile clone
Figure 3‐5 Command line interface (CLI) of the video server generating RTMP video stream
13 Please see http://snap-telemetry.io/ for more information.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 39 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
In the assessment of the mobile cloud use‐cases, several measurements defined and used in the
mobile clone framework to illustrate the benefits obtained, these are listed in Figure 3‐7. In the
video streaming application, the video stream, through the mobile clone, will be transcoded to adapt
to the network status or user devices. The video transcoding will consume a certain amount of CPU
capacity. The in‐going and out‐going video stream of the mobile clone will assume different traffic
volumes as the quality of video differs. Thus, the CPU capacity of the mobile clone and the traffic
through it are monitored. The monitoring of the CPU capacity and traffic can provide the mobile
clone service framework information about the available computation and communication resources
of the mobile clone. The monitoring data can also be used to illustrate the benefits obtained in the
use‐case, for example, if the traffic sent to the user is less than that received by the transcoder, the
video transcoding over the clone saves the bandwidth between clone and user’s mobile device.
Measurements for mobile clone (related to P2 (P2a, P2c), S2)
Definition
CPU Capacity CPU usage of mobile clone in percentile.
Transcoder Input Rate Traffic received from the network in Bytes per second.
Transcoder Output Rate Traffic transmitted to the user in Bytes per second.
Table 3‐6 Measurements for mobile clone
The real‐time telemetry framework Snap used to collect the measurements of this use‐case,
performed the function of VIM in a generic virtualised architecture. Figure 3‐6 shows the surge of
CPU usage of the mobile clone when a user requests a video stream through the video clone which
can provide video transcoding. As was expected, if the transcoding compresses the video stream, the
outgoing network traffic of the clone is less than the incoming traffic, which is shown in Figure 3‐7.
Also, the video proxy permits the live stream (it also could be the transcoded live stream) to be
shared for multiple mobile phones in the same domain, ie, the mobile phones whose clones are in
the same mobile cloud network. The advantage of this is to eliminate the duplicated video traffic
from the video server to the mobile clone network. This sharing functionality has been partly
implemented to show the benefits of a clone as a video proxy; the missing implementation being a
load balancer to automatically forward the video request from its own clone to the neighbouring
clone.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 40 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Figure 3‐6 Monitoring interface of the CPU usage of mobile clone (video proxy)
Figure 3‐7 Monitoring interface of the network interface traffic of mobile clone (video proxy)
Considering the “video streaming in C‐RAN with mobile cloud” application, the results in Table 3‐7
show that the KPI framework is collecting representative KPI values and that the benefits of mobile
clone transcoding are recorded as expected.
KPI for mobile clone Definition Measured performance Comment
CPU Capacity CPU usage of mobile clone in
percentile. Clone CPU usage rises to 25% when transcoding is initiated
The KPI collection framework is providing KPI data as expected
Traffic Received Rate Traffic received of the network interface in Bytes per second.
Received traffic is recorded at ~ 500KB/s
50% reduction in traffic through the mobile clone with video transcoding
Traffic Transmitted Rate
Traffic transmitted of the network interface in Bytes per second.
Transmitted traffic is recorded at ~ 250KB/s
Table 3‐7 KPIs for mobile clone
Considering the 5G business societal and KPIs, the results contribute to the business goal B3
considering the number PoC that demonstrate 5G related technologies, and S3 considering
quantified assessment of maturity and performance of technological components with respect to
cloud technology and complexity and to S4 with respect to stimulation of new economically feasible
applications related to cloud assistance of applications running in user equipment.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 41 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
MobiletaskoffloadinginC‐RANwithmobilecloudIn the mobile task offloading use‐case, the mobile cloud provides the mobile devices with the
capability to offload demanding computational tasks to its corresponding mobile clone in the mobile
cloud. By offloading the computationally‐intensive tasks of the mobile device to the mobile clone,
the energy consumption of the mobile device can be reduced, as well as offering the possibility of
offloading those tasks that exceed the capabilities of mobile device. The components used in this
use‐case are listed in Table 3‐8. The KPI collector differs to the framework used in the video
streaming use‐case, because the mobile clone in this use‐case is running x86 Android OS, which does
not work for the real‐time telemetry framework such as Snap. The VMware vSphere RESTful API was
used to collect the KPIs of the mobile clone via the helper tool vSphere‐influxdb‐go, which runs
periodically using the Linux cron service. The helper tool saved the collected KPIs to the InfluxDB.
Component Description Implementation
Offloading framework – clone service
Serve the task offloading request from mobile phone; execute the offloaded task; send the result of task to mobile phone.
Android native Java code implemented in iCIRRUS
project.
Offloading framework – mobile phone
Offload a task to mobile clone; receive the result of task from mobile clone.
Android native Java code implemented in iCIRRUS
project.
KPI collector Collect the KPI of this use case, such as CPU,
network interface traffic. vSphere‐influxdb‐go14
Linux cron
KPI storage Store the KPI of this use case. InfluxDB
KPI monitoring Publish the KPI of this use case for
monitoring. Grafana
Table 3‐8 Component of the mobile task offloading use‐case
The sequence of steps performed in this use‐case is as follows:
1. The mobile task is sent to the mobile clone on the fly
2. The offloaded task is executed on the mobile clone
3. The result of the offloaded task is sent back to the mobile phone
4. CPU utilization changes on the clone are expected, while offloading requests are being
executed
5. Uplink and downlink network usage changes are observed, as the tasks are offloading and
when the results are sent back to the mobile phone, respectively
In the mobile task offloading application, the reduced energy consumption of a mobile phone is the
major motivation for the use‐case. Thus, the battery information of the mobile phone is collected as
the KPIs of this use‐case, which are illustrated in Table 3‐9. The instantaneous battery current
14 See https://github.com/Oxalide/vsphere-influxdb-go for more information.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 42 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
illustrates the battery drain if the battery is not charging. When a task is executed locally on the
mobile phone, the battery shows significant battery drain. However, if the task is offloaded to the
mobile clone, the battery should show much reduced drainage. The battery voltage is also
monitored so that the instantaneous power consumption can be calculated using the current
multiplied by the voltage. The battery capacity is therefore another important KPI in the task
offloading framework, in which the offloading decision could be made based on the current battery
capacity, which is therefore monitored as well.
Measurements for mobile phone
Definition
Battery Current Instantaneous battery current in microamperes, as an integer. Positive values indicate net current entering the battery from a charge source; negative values indicate net current discharging from the battery
Battery Voltage The present battery voltage level, in volts
Battery Capacity Remaining battery capacity as an integer percentage of total capacity
Table 3‐9 KPIs for mobile phone
In a real‐world environment, the mobile phone state, battery capacity, voltage etc., may be required
by the network infrastructure controller/orchestrator to be used for resource allocation algorithms.
The mobile phone’s battery current, voltage and capacity are all useful in the task offloading use‐
case, so these metrics were collected via the offloading framework installed on the mobile phone.
These collected metrics were saved into InfluxdB and monitored via Grafana, and are shown in
Figure 3‐8 and Figure 3‐9. The energy consumption of the mobile phone was obtained based on the
product of the battery current and voltage. The capacity was only monitored here (ie, not acted
upon); but it could have been potentially integrated into the offloading framework of a decision‐
making algorithm. The computation and network usage of a mobile clone were monitored for
evaluation, and were illustrated in Figure 3‐10 and Figure 3‐11 respectively. When a task was
offloaded to the clone, the battery‐drain of the mobile phone became very minor. On the clone side,
the CPU usages shows significant increase. There was also some network traffic on the clone side
when the offloading request and result were transferred. This use‐case proved that the energy
consumption of a mobile phone could therefore be dramatically reduced by offloading the mobile
task to a clone in the mobile cloud.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 43 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Figure 3‐8 Monitoring interface of the mobile phone battery current
Figure 3‐9 Monitoring interface of the mobile phone's battery voltage and capacity
Figure 3‐10 Monitoring interface of the mobile clone's CPU usage
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 44 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Figure 3‐11 Monitoring interface of the mobile clone's network traffic
Considering the “mobile task offloading in C‐RAN with mobile cloud” application, the results in Table
3‐7 show that the KPI framework is collecting representative KPI values and that the benefits of
mobile clone transcoding are recorded as expected.
KPI for mobile phone Definition Measured performance Comment
Battery Current Instantaneous battery current in microamperes, as an integer.
Positive values indicate net current entering the battery from a charge source, negative values indicate net current discharging from the
battery
Phone current shows activity with a narrow spike but overall
a small impact
The KPI collection framework is providing
KPI as expected
Battery Voltage The current battery voltage level in volt
The battery voltage is displayed
The KPI collection framework is providing KPIs as expected – this value could be used to help support offload
decisions
Battery Capacity Remaining battery capacity as an integer percentage of total
capacity
The battery voltage is displayed
CPU Capacity CPU usage of mobile clone in percentile
Clone CPU usage rises to 15% when application offload is
initiated
The KPI collection framework is providing KPI data as expected
Table 3‐10 KPIs for mobile task off‐loading
Considering the 5G business societal and KPIs, the results contribute to the business goal B3
considering the number PoC that demonstrate 5G related technologies, and S3 considering
quantified assessment of maturity and performance of technological components with respect to
cloud technology and complexity and to S4 with respect to stimulation of new economically feasible
applications related to cloud assistance of applications running in user equipment.
NetworkcontrolledD2Duse‐caseThis Section represents developments to the analysis and simulation results of network controlled
D2D use‐case that were reported in D4.4 [16].
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 45 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
In the network controlled D2D use‐case, the D2D pairs are divided into groups based on their
geographical locations. Each D2D group is covered by one RRS, and the area covering one D2D group
is called a sub‐cell. Orthogonal resource allocation for one D2D group is performed locally at its RRS.
By allowing frequency reuse among different D2D groups or sub‐cells, the spectral efficiency of D2D
communications could be improved by several folds, which contributes the societal goal S2 to
reduce energy consumption. The integration of D2D technology also offers potential for facilitating
development of new applications which support the S4 KPI on service “stimulation.”
Figure 3‐12 demonstrates the spectral efficiency achieved by all D2D pairs when frequency reuse is
allowed among multiple sub‐cells, each of which is covered by one RRS. In this figure, K is the
number of D2D pairs in the system. L represents the number of sub‐cells or D2D groups. Note that
when L is equal to one, there is no frequency reuse among D2D pairs. It can be seen that the highest
spectral efficiency can be 8 times that obtained in the scenario without frequency reuse.
Figure 3‐12 Spectral efficiency achieved by network controlled D2D communication with frequency reuse among L sub‐cells
In iCIRRUS, D2D communication is taken as a key component for reducing the latency of data
transmission. Through simulations, the latency of packet transmission via D2D communications has
been compared with the latency experienced by the packet transmission in traditional cellular
network, in which the traffic will have to go through the BS.
Figure 3‐13 compares the data transmission latency of the D2D communication and that of
traditional LTE communication. In the simulations, both light and heavy traffic load are considered.
Since D2D communication uses a direct link between the source and destination devices, the
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 46 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
transmission latency under D2D communication is about the half of the latency in the LTE system.
The latency gain is even better when the traffic load is high.
Figure 3‐13 Latency Comparison between D2D communications and LTE systems
The key KPI components involved in this use‐case are data rate and latency, which are listed in Table
3‐11.
Network Controlled D2D
KPI Definition
Performance 5G KPI target
D2D data rate
The data rate achieved by
individual D2D pair
Compared with no frequency reuse case,
individual user’s data rate improved significant
when the number of sub‐cells is larger than one.
(P1b) 1‐10Gbit/s
System data rate
The total data rate achieved by all D2D pairs in the coverage of multiple RRHs
controlled by one BBU
Compared with no frequency reuse case, the whole system’s data rate improved significant
when the number of sub‐cells is larger than one
User Latency
The duration for a packet to be
transmitted from the source
(device) to the destination (device).
(P4a) UP latency 1‐10ms
Table 3‐11 D2D Communications performance
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 47 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Ethernettransportlabtestsofnewhigher‐layerfunctionalsplitfronthaul
This Section summarises the test results of the PDCP/RLC higher layer split (HLS) demonstration that
was conducted at Orange’s facility in Lannion and was reported in D5.3 [13], and compares the
results with the iCIRRUS design target KPIs and validates them in the broader context set by the 5G
business, societal and performance KPIs set out in [1]. The analysis of the results in the context of
the overall project is presented in Section 5.
Figure 3‐14 Experimental setup (top) PtP, (bottom) PON shows the experimental set‐up with Point‐
to‐Point (PtP) Ethernet (top) and G‐PON (bottom).
Figure 3‐14 Experimental setup (top) PtP, (bottom) PON
At the transmitter side, generic servers running on OpenStack host a virtual EMS (Element
Management System), a virtual EPC (Evolved Packet Core) where the Mobile Edge Computing (MEC)
was operating, and a virtual BBU (vBBU). A 10G Top of Rack (ToR) Ethernet switch was used to
connect these nodes. At the receiver side, an intelligent RRH (iRRH) was connected to a User
Equipment (UE) (here, a computer) via a mobile dongle. We used a mobile signal based on LTE (Long
Term Evolution) with 15MHz bandwidth and 2x2 MIMO (Multiple Input Multiple Output). The
synchronization in this network is provided by Global Positioning System (GPS) devices connected to
each iRRH.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 48 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Fronthaul KPI PDCP‐RLC split Performance 5G KPI target
User Data rate 100‐400Gbps 0.1Gbit/s (LTE) (P1b) 1‐10Gbit/s
FH Data rate 0.12Gbit/s (LTE) 1
User latency (P4a) 1‐10ms on the air
interface
Max latency (round trip
delay) 60ms 5‐10ms PtP 2
Min frequency accuracy
NA NA
Min phase and timing accuracy
NA +/‐ 900ns downstream +/‐ 48µs upstream
Max latency imbalance
NA +/‐ 40ns downstream+/‐ 140ns upstream
Max BER 10‐3 non‐priority
data3 10‐5 priority data
NA
1. Fronthaul rate is 20% greater than backhaul rate. 2. Dynamic resource allocation in PON adds an extra 1ms to fronthaul latency, avoided with fixed allocation 3. The performance of the HLS experiment was measured at the application layer hence it is appropriate to compare
performance with that required for Ethernet backhaul
Table 3‐12 HLS split at the PDCP‐RLC layer, requirement v performance
The PDCP‐RLC higher layer split generated a fronthaul rate only 20% greater than the user data rate.
This creates an opportunity that existing optical transport deployments using either PtP or PON
architecture may be reused for initial 5G roll‐out. Also, throughput measurements show that
backhaul and FTTH infrastructures can be potentially used for the transport of this Ethernet
fronthaul, albeit with the need to perform adjustments jointly in the RAN and transport equipment
(eg queuing policies).
Considering the 5G performance KPIs P1b and P4a, the results for PDCP‐RLC HLS indicate that
existing fibre infrastructure with a 1G or 10G capability meets, or is close to meeting, the
requirements for per user throughput and latency except for the URLLC use‐case.
Also, considering the 5G business societal and KPIs, the results contribute to the business goal B3
considering the number of PoC that demonstrate 5G related technologies, and S3 considering
quantified assessment of maturity and performance of technological components, and especially
considering estimation of the TCO and roll‐out that would be greatly assisted by the potential to re‐
use existing fibre infrastructure. Additionally, as the RAN is software based with a vEPC, vBBU and
MEC platform, hosted on servers based on OpenStack the demonstration contributes to the 5G
performance goal P2a considering automation of service creation and P2c considering use of
opensource SDK.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 49 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
4 DescriptionandevaluationofthefinalShowcasedemonstrator
DescriptionoftheShowcasedemonstratorThe overall setup of the demo testbed is depicted in Figure 4‐1, where five different front and
backhaul services were successfully aggregated and converged onto a single metro fibre link. Two
identical 100G aggregators were deployed at the CO and remote node. The metro link used in the
showcase demonstration test is shown in Figure 4‐2. It consists of an average length optical span
(3km) which is used for backhauling mobile connections and for C‐RAN fronthaul, and can cover the
range of a city C‐RAN cluster [D2.3] [6]. In the actual test scenario, the link was looped at the
terminal BS location, resulting in a doubled effective optical path of 6km. In this case, a typical
network link length was emulated, which introduced an additional 30μs of transport delay and
represents a realistic scenario for the equipment under test.
Figure 4‐1 iCIRRUS testbed setup at Telekom Slovenije
At the CO, the legacy CPRI (Option 3) traffic was generated by the VIAVI MTS‐5800 tester, and the
CPRI streams were then de‐mapped and packetized in the normal Ethernet frames. The size of an
Ethernet packet can be configured by software, while each data section of a MAC packet consists of
an internal header information field and 2n CPRI basic‐frames. Typical configurations used for the
tests were payloads of 4, 8 and 16 basic‐frames per MAC‐packet for 2.45Gbit CPRI and 4 basic‐
frames per MAC‐packet for 9.83Gbit CPRI. With these configurations, the effective packet‐sizes can
be held within the standard MTU size of 1492 bytes. The effective packet‐sizes can be found in Table
4‐1.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 50 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
CPRI 2.45Gbit CPRI 9.83Gbit
Number of basic‐frames
Size (byte) Ratio Header/Packet
Size (byte) Ratio Header/Packet
1 80 20.00% 256 6.25%
2 144 11.11% 496 3.23%
4 272 5.88% 976 1.64%
8 528 3.03% 1936 0.83%
16 1040 1.54% 3856 0.41%
32 2064 0.78% 7696 0.21%
64 4112 0.39% 15376 0.10%
128 8208 0.19% 30736 0.05%
256 16400 0.10% 61456 0.03% Table 4‐1CPRI over Ethernet ‐ MAC payload‐sizes
On the second client port, an Ethernet stream with 9Gbit/s traffic load and a fixed frame size of 1518
Bytes was generated by the VIAVI ONT‐606 tester. The 60GHz UPPER PHY LLS was connected to the
third 10G client port. Incoming Ethernet frames were processed by the DU, and the fronthaul data
was passed to the aggregator. Due to physical layer limitations, the data rate for the 60‐GHz link was
limited to a gross rate of 2.5Gb/s. A flow control protocol between DU and RU triggered a
backpressure mechanism back to the data source, to avoid overflow of the fronthaul system.
On the fourth 10G client port, a software‐based base station (BS) running the SRS‐LTE software15 was
connected. Specifically, one port of the BS was connected to the aggregator access port through a
10GbE switch while a second port was connected to the backhaul (Evolved Packet Core, EPC). The
proof‐of‐concept application for cloud computing were also hosted over this link.
The last client port was connected to the TS core network via 10G Ethernet router, to provision the
4G‐LTE femto‐cell access at the remote node. To enable the synchronization between the CO and
RN, an ADVA OSA‐5410 was configured as a grand master clock in accordance with the ITU‐T
G8275.1 PTP profile, receiving the reference from GNSS satellites and establishing the PTP link via
the 1G client port on the aggregator. The 10MHz clock output was used to trigger the CPRI traffic
generation. The PTP packets were then multiplexed with higher priority over other SM traffic and
broadcast to all the client ports of the remote aggregator, more details can be found in D5.3 [13].
15 SRS-ENB, Software Radio Systems, http://www.softwareradiosystems.com/
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 51 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Figure 4‐2 Metro fibre link (in red) used in the iCIRRUS integration demo
After aggregation at the 100G interface and transmission over 6km of TS metro fibre, different traffic
streams were disaggregated on the client side accordingly. On the 1G client port, another ADVA
OSA‐5410 terminated the PTP packets as a slave clock, of which both time and phase could be
synchronized to the grand master clock. The internal 10MHz clock was also locked to the recovered
PTP clock. The CPRIoEth transponder board was externally triggered by this 10MHz clock, to recover
the accurate CPRI data clock and depacketize the original CPRI stream. The recovered CPRI data was
then analysed by the VIAVI MTS‐5800. Similarly, the 9Gbit/s Ethernet traffic was looped back on the
client port to the CO for analysis. The 60‐GHz RU was connected to the third client port. The RU
processes the fronthaul data and generates the physical layer information. A 60GHz link was
connected to the RU downstream port in a loopback configuration. A commercial remote radio
head, used to transmit the LTE signal (generated by the software BS) over‐the‐air, was connected in
the fourth client port, with the LTE signal being received by a commercial UE phone (used during
cloud computing experiments).
The following Sections outline the adaptation to the individual testbeds undertaken to enable their
incorporation into the Showcase demonstrator. Then present the results obtained for each of the
components, under simultaneous operation. Finally, an evaluation of the Show case demonstrator is
provided that highlights the results obtained in the wider context of 5GPP business, societal and
performance KPIs.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 52 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
CPRIoverEthernettestandverificationintheShowcaseThe individual testbed was described in Section 3.1 together with the test and verification of its
associated KPIs.
This Section presents the necessary adaptations the testbed to enable integration and results of the
test and verification of CPRI over Ethernet in the Showcase demonstrator.
For the 100G aggregator scenario (Table 4‐7), the frequency synchronisation mechanism slightly
different to that in Section 3.1 is required. Because the 100G aggregator did not support
synchronous Ethernet operation, the frequency adjustment had to be realized using ADVA’s OSA‐
5410 device, as already described in Section 3.1. From the FPU and clock distribution’s point of view,
only the RRH side software configuration had to be changed. The on‐board DSPLL was re‐
programmed to generate its internal CPRI and Ethernet Core clocks from the provided OSA reference
clock. It turned out that both CPRI clocks at the BBU and RRH sides could be effectively synchronized
via the highly prioritized PTP link. A frequency offset between the BBU side and RRH side could not
be determined; the requirements of Table 2‐2 demanding a minimum frequency accuracy of ±2ppb
were therefore met.
iCirrus FPGA Board ‚BBU Side‘
VIAVI MTS 5800CPRI Tester
CPRI Master
Port
CPRI / Ethernet BridgeClock Unit(DSPLL)
RecoveredCLK CPRI
REF CLK
Eth REF CLK
CPRI Core
10G Eth Core
CPRI SlavePort
BERMeasurement
iCirrus FPGA Board ‚RRH Side‘
CPRI / Ethernet Bridge
Clock Unit(DSPLL)
ReferenceCLK
CPRIREF CLK
Eth REF CLK
CPRI Core
10G Eth Core
CPRI CPRI
10G Eth
OSA OSAGPS Reference
Aggregator10GEth
1G Eth
100GEth
1G Eth. PtP Sync Packets
10 MHz Reference CLK
Aggregator10GEth
1G Eth
100GEth100G Eth
1G Eth. PtP Sync Packets
10G Eth
Figure 4‐3 CPRI/Ethernet bridge connected to aggregator without SyncE
As CPRI is a continuous data stream, every unexpected delay of incoming data would result in a total
failure of the CPRI link. Because of the number of sources connected to the aggregator and the
transmission of prioritized services, the Ethernet packets of the CPRI link will sometimes be affected
with temporary increased latency (packet delay variation). To improve the stability of the CPRI link
within the 100G network, additional buffer FIFOs were added to both CPRI RX data paths at the BBU
and RRH sides. At start‐up, the “playout” of CPRI data on the RX side will not start until a software
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 53 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
pre‐programmed buffer level is reached. These buffers introduce an additional static latency from
the CPRI Master to the CPRI Slave side, but prevent the RX side from running out of data when a
subsequent packet arrives with an unexpected delay offset. The additional static delay correlates
with the buffer level and the size of Ethernet MAC packets containing the CPRI data. Figure 4‐4 and
Figure 4‐5 show the measured one‐way latency (VIAVI CPRI master port to VIAVI CPRI slave port)
with FIFI buffer levels “2.45Gbit CPRI” and “9.83Gbit CPRI” respectively. The internal latency within
the VIAVI MTS5800 and the latency of the 100G aggregator have been eliminated.
Figure 4‐4 Latency vs. FIFO buffer level (2.45Gbit CPRI)
Figure 4‐5 Latency vs. FIFO buffer level (9.83Gbit CPRI)
The latency displayed at FIFO Fill Level “0” represents the system’s internal signal path delay through
all the involved CPRI and Ethernet cores and all the additional FPGA firmware‐modules (basic one‐
way‐latency). The 2.45Gbit mode is affected with a much higher basic one‐way‐latency (~17.5 µs)
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 54 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
than the 9.83Gbit mode, which is the result of higher CPRI core latencies and a higher internal data
acquisition time. The 2.45Gbit CPRI can therefore be taken as the worst‐case mode.
The tests showed that a small RX buffer‐level adding < 10µs of latency was enough to prevent the
Ethernet‐to‐CPRI module from running out of data. For the 2.45Gbit mode, the resulting delay has to
be increased by ~10µs to an overall one‐way‐latency of ~ 27.5µs. Because of the symmetric design of
the CPRI‐over‐Ethernet firmware, the round‐trip‐delay can be calculated as
~ 27.5µs*2 = ~55µs, which covers the requirements of Table 2‐2.
A comparison with the target KPIs is provided in Table 4‐2 and Table 4‐3 the Tables also include the
standalone performance reported in Section 3.1.
Fronthaul KPI Requirement Measured
performance Comment
Data rate 100‐400Gbps ‐
Max latency (round trip
delay)
150µs (CoMP) 440µs (no CoMP)
35µs direct link 55µs aggregated
link
Exceeds most demanding requirement
Min frequency accuracy
+/‐ 2ppb (per hop)
Better than (± 4.92Hz
@2.45Gbit)
Actual error was less than measurement capability in with direct connection and via
aggregated link
Min phase and timing accuracy
+/‐ 10ns (MIMO & TX diversity)
Aggregated link 30ns average, 1460ns max
Measured on aggregated link
Max latency imbalance
+/‐ 16ns Not measured
Max BER Below measurement sensitivity
Table 4‐2 Legacy 2.45Gbit CPRI traffic over Ethernet, requirement v performance
Fronthaul KPI Requirement Measured
performance Comment
Data rate 100‐400Gbps ‐
Max latency (round trip
delay)
150µs (CoMP) 440µs (no CoMP)
10µs direct link30µs aggregated
link
Buffer depth set by more latency consuming 2.45Gbit
Min frequency accuracy
+/‐ 2ppb (per hop)
Better than (± 19.66Hz
@9.83Gbit)
Actual error was less than measurement capability in with direct connection and via
aggregated link
Min phase and timing accuracy
+/‐ 10ns (MIMO & TX diversity)
Aggregated link 30ns average, 1460ns max
Measured on aggregated link
Max latency imbalance
+/‐ 16ns ‐ Not measured
Max BER ‐ Below measurement sensitivity
Table 4‐3 Legacy 9.83Gbit CPRI traffic over Ethernet, requirement v performance
The results indicate that CPRI over Ethernet can exceed most of the KPI requirements that were
identified. However, the peak timing accuracy performance as measured on the aggregated link
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 55 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
indicates that performance may not be sufficient for the most severe MIMO and TX diversity
requirements.
Extending what was stated in Section 3.1, demonstration of CPRI over Ethernet over an aggregated
fronthaul can be related to the business and societal KPI B3/S3 related to cost of ownership and
rollout of 5G with its support of backward compatibility.
Real‐timefronthaulplatformusing60GHzwirelesswithUPPER‐PHYLLSverificationintheShowcase
This Section describes the test results of the 60‐GHz fronthaul system, which have been gained
during the integrated trial in Ljubljana at the TS labs. A description and the measurement results of
the standalone 60‐GHz fronthaul platform can be found in D5.3 [13]. Due to export regulations by US
authorities, it was not possible to ship the UE device to the final test. Therefore, a loopback
configuration at the RU has been used. A block diagram of the tested 60‐GHz fronthaul system can
be found in Figure 4‐6.
Figure 4‐6: 60‐GHz fronthaul system with radio loopback (Ethernet analyser shown as data source and sink)
Referencemeasurement(fronthaullinkonly)In a first series of measurements, the fronthaul link was characterized.
Two general trends can be observed for the throughput measurement. First, the L1 throughput
decreases with increasing frame size. This can be explained with the control mechanism that
guarantee that at the 60GHz physical layer only transmits 2.5Gb/s. Incoming Ethernet frames will be
stripped of Preamble and FCS and forwarded via the fronthaul link to the 60GHz PHY. Since the PHY
rate at the 60GHz is always 2.5Gb/s, for short Ethernet frames more preambles and FCS information
will be put into the system and therefore a higher L1 rate is required.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 56 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Figure 4‐7: Throughput for 60GHz fronthaul link for different frame sizes at different layer 1…4 (L1…L2)
Second, for Layer 2 and above the opposite effect can be observed. Shorter Ethernet frames (or IP or
TCP packets) have more overhead which reduces the net data rate.
The analysis of the system latency has confirmed the previous results of latency values in the order
of 700µs. The jitter was measured with values between five and eight µs. Both results are shown in
Figure 4‐8 and Figure 4‐9.
Figure 4‐8: Latency vs. frame size for 60GHz fronthaul link
Figure 4‐9: Jitter vs. frame size for 60GHz fronthaul link
Inclusionofaggregatorswitchesinto60GHzfronthaullinkIn a next step, the two aggregator switches have been inserted into the link (Figure 4‐10) and the
measurements have been repeated. The aggregator nodes have also carried other traffic, eg PTP
traffic.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 57 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Figure 4‐10: 60GHz fronthaul system with radio loopback and two aggregator nodes
The throughput, latency and jitter measurements as shown in Figure 4‐11, Figure 4‐12, and Figure
4‐13 do not show any significant difference compared to the reference results discussed in Section
4.3.1. It must be noted that the latency is not increased, although two additional elements have
been integrated into the system. The latency is determined by the flow control algorithm, which
requires an input queue of a sufficient depth to avoid jitter in the output data.
Figure 4‐11: Throughput for 60GHz fronthaul link with aggregator switches for different frame sizes at different layer 1…4 (L1…L2)
Figure 4‐12: Latency for 60GHz fronthaul link with aggregator switches for different frame sizes
Figure 4‐13: Jitter (ave) for 60GHz fronthaul link with aggregator switches for different frame sizes
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 58 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
The measured latency figures of ~700µs are above the KPI requirement as discussed in Table 2‐2.
This can be explained by the requirement of the flow control algorithm, which adjusts the source
data rate to avoid an overflow of the 60GHz physical layer, which is limited to 2.5Gb/s currently.
Currently, the data source can send out bursts of data frames with a rate of 10Gb/s. The last frames
of such a burst must be delayed most, which causes this relatively high latency. The latency can be
reduced if the data source would send out data with a more even distribution.
Inclusionof3kminstalledfibreintofronthaullinkThe inclusion of a 3km fibre link into the testbed triggered an error in the fronthaul system, which
could not be fixed so far. To investigate this issue, the previous setup as shown in Figure 4‐10 was
modified insofar that the aggregation nodes have been removed, but the short link between DU and
RU was increased to a length of 3km now. Consequently, the measurement setup is very like the one
shown in Figure 4‐6 (except for the 3km link between DU and RU).
In this configuration, the following results for throughput, latency and jitter have been measured
(Figure 4‐14, Figure 4‐15, Figure 4‐16).
Figure 4‐14: Throughput for 60GHz fronthaul link with 3km fibre for different frame sizes at different layer 1…4 (L1…L2)
Figure 4‐15: Latency for 60GHz fronthaul link with 3km reach for different frame sizes
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 59 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Figure 4‐16: Jitter for 60GHz fronthaul link with 3km reach for different frame sizes
When introducing a 3km link between DU and RU a loss of frames could be observed. The reason for
this behaviour can be again explained with the flow control algorithm. It works in the following way.
If the receiver of an Ethernet link gets more frames than it can process (detected by a threshold in
the input queue fill state), it sends out a PAUSE frame in direction of the data source. The source
should then stop the data traffic for the duration indicated in the PAUSE frame. If the algorithm
works correctly in a fully loaded system this means we get bursts of frames with a line rate of 10Gb/s
and a period with no data. In the case of a targeted rate of 2.5Gb/s the burst duration should be ¼ of
the time gap between burst.
If the link between DU and RU is long, there are many frames “stored” in the fibre. If the PAUSE
frame reaches the data source all frames stored in the fibre still reach the receiver. If the storage
capacity of the fibre and input queue are not correctly balanced and frames get dropped.
It is planned to modify the queue size and the thresholds for a final test to avoid the loss of data
frames.
Fronthaul KPI
Requirement Upper PHY split in DL and UL (no
CoMP)
Requirement Upper PHY split in DL lower PHY
split in UL (CoMP)
Performance 5G KPI target
User Data rate 100‐400Gbps 100‐400Gbps 2.2Gb/s (P1b) 1‐10GB/s
FH Data Rate
User Latency (P4a) 1‐10ms on the air
interface
Max latency (round trip
delay) 440µs 150µs 800µs
Min frequency accuracy
+/‐ 2ppb (per hop)
+/‐ 2ppb(per hop)
Min phase and timing accuracy
+/‐ 30ns +/‐ 30ns
Max latency imbalance
+/‐ 163ns +/‐ 163ns
Max BER 10‐12
Table 4‐4 lower layer split at the UPPER PHY, requirement v performance
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 60 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
The results of the FPGA‐based DU radio in section 3.3 indicate that the described fronthaul solution
falls outside the latency requirements, which indicates that further optimisation is required. The
major reason for the latency in the order of 800µs is the queuing required for the flow control. The
observed jitter can also be explained by the flow control mechanism. The 60GHz wireless link also
operates using a low latency scheduler, however, due to legal issues the 60GHz mobile, the link was
operated in a radio‐loop back mode and representative performance data was not available.
Software‐based‐DUradiooverEthernetwithMAC‐PHYLLSvalidationintheShowcase
The set‐up for the functional split case and the generated packet types are described in detail in [17]
and [5], while the analysis presented here is based on the results from [18]. The validation set‐up is
shown in Figure 4‐17, together with the different interface points where measurement results can
be extracted. The generated packet types include the Physical Downlink Shared Channel (PDSCH),
Downlink Control Information (DCI), Random Access Response (RAR) and System Information (SI)
packet type. For these tests, the maximum packet size is for the DLSCH packet type and it is
approximately 1000 Octets long, but this value is shared by the number of recipients (users) within
the subframe. For these tests, a maximum number of 3 users were active within a subframe at any
given time. The transport network operates at 1Gbps (1GbE) and includes only Layer‐2 processing.
The results presented here include the mean and standard deviation (STD) of packet latency, which
is defined Table 2‐2 (albeit as two‐way latency, while the measurements presented here are for one‐
way latency), and subframe (LTE) latency. The latter is important as it relates to the length of time
that Ethernet packets that form LTE subframes must be queued at the RU prior to playout of the
subframe.
Figure 4‐17 Set‐up used for validation testing (left) and interface measurement points (right)
The packet latency was measured between interface points 1 and 3. It is an average of the delay
encountered by the 4 different packet types during the transmission of 9000 LTE subframes. The
subframe latency took into account the processing time of each subframe, and was measured at the
same interface points (1 & 3) but taking into account the first packet in each subframe (a DCI packet‐
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 61 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
type) at interface point 1 and the last packet for that same subframe (PDSCH, SI or RAR packet types)
at interface point 3. The results of Figure 4‐18 included the contention effects with bursty
background traffic of different burst sizes and different packet lengths, while for those in Figure
4‐19, the background traffic is constant‐packet rate (CPR). In both sets of results, the baseline case is
without contention. Table 4‐5 shows a summary of the average latency for the different networking
configurations, while Table 4‐6 extrapolates these results for a network with 10GbE.
Figure 4‐18 Fronthaul latency with two Ethernet switches over a 1Gbps Ethernet network, Mean (Left) and Standard Deviation (Right). The traces include packet latency (“per‐packet”) and subframe latency (“per‐subframe). The dotted lines
are baseline cases while the rest of the traces are with bursty background traffic.
Figure 4‐19 Mean and standard deviation for the fronthaul latency with two Ethernet switches over a 1Gbps Ethernet network. The traces include packet latency (“per‐packet”) and subframe latency (“per‐subframe). The dotted lines are
baseline cases while the rest of the traces are with constant packet‐rate background traffic.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 62 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
KPI No Switch 1 Switch 2 Switches
Packet latency mean (µs) 45 53 60
Subframe latency mean (µs) 54 62 69
Packet latency STD (µs) 9 N/A N/A
Subframe latency STD (µs) 6.3 N/A N/A Table 4‐5 Measurement results for the packet and subframe latencies for the baseline cases (no contention) (x indicated not
measured)
KPI No Switch 1 Switch 2 Switches
Packet latency mean (µs) 43 48.2 53.5
Subframe latency mean (µs) 52 57.3 62.5
Packet latency STD (µs) 6.9 N/A N/A
Subframe latency STD (µs) 4.1 N/A N/A Table 4‐6 Estimated packet and subframe latencies for the baseline case (no contention) for 10GbE networking (x indicates
not measured)
Based on these results, an extrapolation of the latency for the set‐up with the iCIRRUS aggregator
can be carried out in a crude manner as follows: Assuming 10GbE access, the standard deviation will
be dominated by the processing delay in the DU, including consideration of a one‐way delay of 35 s for the aggregator. The result of this extrapolation is shown Table 4‐7. The measurement results
with contention can also be similarly extrapolated. However, for large aggregation levels the effect
of contention is insignificant. Based on the results in Figure 4‐18 and Figure 4‐19, for large
aggregation levels, for example, a ratio of split traffic data to total aggregated data of 1.25%
(assuming 80% utilisation of the aggregator link rate, this would correspond to a split data rate of
1Gbps) and a packet size of 1024 Octets (for the background traffic), the increase in packet latency
and STD will be insignificant.
KPI iCIRRUS Aggregator
Packet latency mean (µs) 78
Packet latency STD (µs) 6.9
Subframe latency mean (µs) 87
Subframe latency STD (µs) 4.1 Table 4‐7 Mean and standard deviation (STD) of the packet and subframe latencies with the iCIRRUS aggregation set‐up.
The values are extrapolated from the measured results.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 63 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Fronthaul KPI
Requirement Upper PHY split in DL and UL (no
CoMP)
Requirement Upper PHY split in DL lower PHY
split in UL (CoMP)
Extrapolated performance
Data rate 100‐400Gbps 100‐400Gbps 1Gbps
Max latency (round trip
delay) 440µs 150µs 156µs
Min frequency accuracy
+/‐ 2ppb (per hop)
+/‐ 2ppb (per hop)
N/A
Min phase and timing accuracy
+/‐ 30ns +/‐ 30ns N/A
Max latency imbalance
+/‐ 163ns +/‐ 163ns N/A
Max BER 10‐12 10‐12
Table 4‐8 MAC‐PHY LLS, requirement v performance
The results of the software‐based‐DU radio in Section 3.3 indicate that the MAC‐PHY LLS only slightly
exceeds the most stringent RTT latency requirements that were identified, which indicates that the
majority of CoMP approaches would be supportable. A large percentage of the quoted one‐way
latency (78 µs) is fixed by the aggregator latency while the rest is mainly due to the software
generation process in the end station. The latter has the potential of significant reduction with
hardware acceleration technologies. The remaining quoted KPI values in Table 4‐7 cannot be
compared to specific requirements as such requirements do not currently exist. The reason is that
the subframe‐level KPI requirements will be implementation dependent. That is, they will vary from
one implementation to the next and will need to be absorbed through buffering at the end stations.
It is also important to note, that the packet jitter KPI that is defined in Table 4‐8, does not apply
here. The reason is that with the iCIRRUS aggregator the PTP traffic is not contending with the
MAC/PHY split traffic. However, the case is made here that operators will need to monitor these
additional KPIs (subframe‐related) and constrain them to acceptable levels based on their specific
architecture requirements.
In addition to the societal KPIs that were discussed in Section 2.1, regarding the importance of the
evolved fronthaul for meeting these KPIs, several performance KPIs were also defined. The MAC‐PHY
LLS evolved fronthaul can be assessed in terms of adherence to these KPIs, in an indirect‐
quantitative manner (by extrapolating from benchmarking results) and in a qualitative manner.
Referring to Figure 4‐20, the data rates over the LLS fronthaul can be seen to scale with the cell load.
This is an important result when considering 5G use cases with increased data rate demands.
Extrapolating from the results shown here, assuming a 10GbE link and same levels of overhead (note
that these are based on 4G, optimisations may be expected for 5G), the LLS can potentially support
the user throughputs defined in KPI P1a, something that would be extremely challenging to achieve
with current transport technologies (generic IQ, CPRI). Note that benchmarking tests (summarised in
Table 3‐4) with software‐based DU implementations have shown that they can scale to the levels of
user throughput allowed by current 4G bandwidths. Similarly, with dense RU deployments, it will
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 64 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
potentially be possible to offer the traffic levels expected by KPI P1a. Furthermore, the efficient use
of available transport bandwidths (compared to current fronthaul transport technologies), means
that the system has the potential of meeting KPI P3b.
While the extrapolated latency values presented in Table 4‐7 do not consider the end‐to‐end latency
(that is, including the processing in user equipment), the latency contributions that are taken into
account are such that they should allow the system to meet KPI P4a.
The LLS is a software implementation, running on open‐source software modules and off‐the‐shelf
hardware (software‐defined radios and Ethernet equipment). As such it meets KPI P2c.
Currently the LLS implementation does not meet KPI P4b specifications. However, the system is
currently solely software‐based and it is expected that significant reductions in packet latency
variation can be achieved through hardware acceleration techniques.
The LLS fronthaul implementation has been presented in several peer‐reviewed conferences. The
initial benchmarking results were first presented in [10]. The performance of the system under
traffic contention was presented in an invited paper [14], while first experimental results using strict
priority scheduling in the LLS fronthaul were presented in [11]. There are further plans to improve
and optimise the system while its contribution to new research and innovation calls (national and
EU‐based) is considered as significant.
Figure 4‐20 Benchmarking results for the LLS evolved fronthaul. The fronthaul data rate scales with the application data rate (equivalent to the cell load) while the overhead remains approximately constant at around 42%.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 65 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Femto‐cellfronthaul/backhaulaggregationvalidationintheShowcase
Small/femto‐cell (eg indoor, home cell) deployment strategy is a general mobile operator strategy to
extend the network coverage to indoor and network edge. In the 5G fronthaul concept considered in
the iCIRRUS project, the high capacity network infrastructure could be shared with the fixed internet
access. In this case it can be used as a high quality backhaul for small cell deployment.
As was confirmed in the integration demo test, the small cell backhaul traffic could be efficiently
integrated in the 100G fronthaul segment as low priority traffic. It is based on an IPsec connection to
the mobile network core, which provides transparent, interoperable, and cryptography‐based
security services to ensure confidentiality, integrity, and authenticity of data and to provide anti‐
replay protection. The system itself provides automatic cell provisioning without operator needs for
manual configurations. Due to the limited required system (intercell) coordination, for handover
only (no CoMP), the backhaul performance is not critical (eg latency). For proper small cell operation
and good user experience, enough backhaul bandwidth and low latency is required, which is assured
by the 100G trunk. The measured 0.07 ms RTT of the ADVA aggregator switch is well below the most
severe upper limits values (1ms) for the backhaul as defined in the previous D2.1 [2] report.
In the test, a Cat 4 LTE femto‐cell with nominal values (300Mbps DL/50Mbps UL) was used. In our
speed‐test we measured 110Mbps DL, 42Mbps UL, and 30 ms RTT (ping). We can therefore conclude
that the traffic was not limited by the ADVA aggregator switch, because we measured practically the
same values without it.
The benefit of using a high capacity fronthaul network for small cell backhaul is the high‐grade
service QoS due to the carrier grade network segment used in the test. It is entirely under the
service provider’s management. In contrast, in the case of a customer fixed access network as the
small cell backhaul, the small cell QoS can vary depending on the fixed‐line personal use, and so is
not under the control of the mobile network operator.
Fronthaul KPI Requirement Measured
performance Comment
Data rate 100‐400Gbps
Max latency (round trip
delay) 60ms
30ms(0.07ms
aggregator performance)
RTT delay meets requirements,
not influenced by aggregator
Min frequency accuracy
N/A N/A
Min phase and timing accuracy
N/A N/A
Max latency imbalance
N/A N/A
Max BER 10‐3 non‐priority
data No errors measured
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 66 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
10‐5 priority data
Table 4‐9 Backhaul (unified transport), requirement v performance
The experiment showed that the aggregated link could seamlessly support integration of backhaul
traffic. Performance degradation due to the inclusion of the aggregator was found to be negligible.
The result supports the business and societal KPI B3/S3 goal considering 5G enabling PoC and also in
understanding TCO the performance of RG enabling technology components. Additionally, the
aggregation of multiple 10G streams onto a 100G fronthaul link with TSM support for PTP operation
is also enabling for the performance 5G KPIs including P1b per user throughput, and P4a on latency.
AggregatedEthernet‐basedfronthaulnetworkvalidationintheShowcase
As depicted in Figure 4‐1, a 10GbE traffic flow was generated by a VIAVI network tester at the local
aggregator and looped back from the remote aggregator to evaluate the round‐trip transmission
performance of the fronthaul link. The traffic profile was configured with a fixed frame size of 1518
Bytes and a sustained a throughput of 9Gbit/s, and the transmission was kept running for 60 hours
without any bit error and packet loss. Table 2‐2 and Table 4‐6 summarise the measured round‐trip
transfer delay and packet jitter.
Average (µs) Minimum (µs) Maximum (µs) Variation (µs)
69.712 69.610 71.360 1.75 Table 2‐2 Round‐trip time delay of 10GbE traffic flow
Average (ns) Minimum (ns) Peak (ns)
30 0 1460 Table 2‐3 Instantaneous jitter of 10GbE traffic flow
It is worth pointing out that during the test, the PTP flow was always broadcast at the same time as
the GST class, which led to a higher delay variation and packet jitter on the statistically multiplexed
(SM) traffic (ie 10GbE). Similar results can be also found in the back‐to‐back case in D5.3, Section 2.5
[13]. In addition, compared to the back‐to‐back result, the 6km metro fibre link introduced another
round‐trip propagation delay of 63.2µs.
The performance comparison of different Ethernet traffic patterns can be found in D5.3, Section 2.5
[13]. There was no additional impact observed in the field integration.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 67 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Fronthaul Requirements
Legacy traffic (CPRI)
Upper PHY split in DL and UL (no CoMP)
Upper PHY split in DL
lower PHY split in UL (CoMP)
PDCP‐RLC split Backhaul (unified
transport)
Measured performance
Data rate 100‐400Gbps 100Gbps per network
(trunk) port
FH data rate 10 Gbps per
RRH
Max latency (round trip
delay)
150µs (CoMP)
440µs (no CoMP)
440µs 150µs 60ms 60ms
6.5µs (2 nodes) +
approx.. 10µs per fibre km
Max BER 10‐12 10‐6
10‐3 non‐priority data 10‐5 priority
data
Error‐free
Table 4‐10 Fronthaul data plane requirement v performance
Table 4‐10 summarises the measured results in comparison with the fronthaul KPI. Considering the
most stringent latency requirement stated in Table 2‐2, which is a max round‐trip delay of 150µs in
case of enabling CoMP in the legacy CPRI link, the proposed fronthaul network is still able to
aggregate 10 times 10GbE traffics over a single 100G trunk link of up to 14.6 km distance. This
provides further support that the aggregation testbed is enabling for the performance 5G KPIs
including P1b per user throughput, and P4a on latency
EvaluationoftheShowcasedemonstratorThe aim of this Section is to draw together contributions that have been made towards validating
the iCIRRUS project’s outputs in the context set by the 5G business, societal and performance KPIs
set out in [1].
The Showcase simultaneously demonstrated many of the iCIRRUS use‐cases over a common
infrastructure in the laboratories of Telekom Slovenije. The Unified Communication use‐case also
demonstrated integration with cloud servers located at Telekom Slovenije.
As set out in the preceding subsections, the following technologies were incorporated, and
simultaneously operated in the Showcase,
CPRI over Ethernet (Section 4.2)
60GHz wireless with UPPER‐PHY LLS (Section 4.3)
Software‐based‐DU radio with MAC/PHY LLS (Section 4.4)
Common transport with femto‐cell fronthaul/backhaul aggregation (Section 4.5)
100G TSN Ethernet aggregated fronthaul (Section 4.6)
Unified Communication video application with mobile cloud support (Section 3.5)
Unified Communication application offload to mobile cloud (Section 3.6)
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 68 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Other technologies were evaluated independently of the Showcase. However, it is useful to consider
the KPIs validated by these experiments to provide a thorough overview of the project.
Network controlled D2D simulation analysis (Section 3.7)
vRAN demonstration with PDCP/RLC HLS (Section 3.8)
The 5G KPIs are addressed in a sequential fashion with a short description of the contribution that
each element makes, but with B3 and S3 being dealt together due to the similarity of their
underlying objective,
B3: “Reach a global market share for 5G equipment and services delivered …,” with proxy
metrics as number of PoC developed and disseminated, contribution to standards and IPR
development and S3: “European availability of a competitive industrial offer for 5G systems
and technologies,” with proxy metrics, quantified assessments of maturity and performance
of technological components, of estimating TCO and rollout.
There were many significant contributions to KPI B3/S3. Foremost, is the Showcase demo itself as a
PoC that demonstrated many 5G enabling technologies with simultaneous operation of most of the
testbeds above. As described in the Sections for each individual testbed B3/S3 is supported in a
similar way by each. Also, considering the contribution to support of innovation in the wider
community, as set out above and detailed in D6.4 [19], many members attended standardisation
bodies representing iCIRRUS views and several iCIRRUS specific contributions, 5 patents were filed
and over 20 publications and presentations were made. Estimates of the impact of various
technologies on TCO and rollout were also made, for example, simulation and analysis in the case of
D2D and extrapolation from field data in the case of C‐RAN, see [6] for details. Additionally,
considering the initial roll‐out of 5G the PDCP/RLC HLS demo shows existing 1G and 10G transport
infrastructure meets requirements for providing support for TCO and initial roll‐out.
S2: “Reduction of energy consumption per service up 90%.”
Contributions to the S2, reducing energy consumption per service, are made by the cloud‐supported
Unified Communication applications that demonstrate reduction of energy consumption in mobile
devices and efficient processing of application data in the mobile cloud, and D2D communication
where enhanced spectral efficiency supports use of lower transmit powers. It is also made by the
vRAN/C‐RAN iCIRRUS architecture itself, and by the use of lower bit‐rate transmission in the
fronthaul, enabled by the RAN functional splitting, and the iCIRRUS Ethernet transport profiles, and
efficient, time‐sensitive multiplexing and aggregation which support these.
S4: “Stimulation of new economically viable services …”
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 69 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Contributions to S4 were made by made by several testbeds including mobile cloud assisted
applications, enabling applications potentially inhibited by user equipment power and energy
considerations to be addressed, 60GHz wireless facilitated by opening up new wider bandwidths,
and network controlled D2D communication which facilitates new ways of delivering applications.
P1: Providing 1000 times higher wireless area capacity …” with proxy metrics of absorbing
P1a 10TB/s/km2, and P1b reach a peak user data rate of 1‐10Gbit/s.
Contributions to P1b are made by D2D where simulation confirms that short range between D2D
devices and enhanced spectral efficiency promote user data rate, and 60GHz wireless with UPPER
PHY LLS that demonstrates a 2.2Gbit/s user rate. The transport for new RAN functional split
interfaces also facilitates these KPIs, as (1) fronthaul bit rates become closer to user data rates,
instead of more than an order of magnitude greater as in technologies such as CPRI, and (2) the
aggregation levels demonstrated at 100 Gb/s will enable interconnection of numbers of mmW cells
in dense or office areas which permit the attainment of 10 Tb/s/km2 area capacity.
P2: “Reducing average service creation time…” with proxy metrics of P2a level of automation
of service creation, P2b existence of autonomic management framework and, P2c, usage of
opensource SDK.
Contributions to the P2a and P2c KPIs are made by the cloud assisted Universal Communication
applications, which relate to automated service creation and use of opensource software.
Additionally, both the MAC/PHY LLS and the PDCP/RLC HLS contribute to P2a considering RAN‐as‐a‐
service, and to P2c as they are based on opensource software.
P3: “Facilitate very dense deployments…” with proxy metrics handle P3a handling 10,000 to
1M devices/km2, P3b facilitate large numbers of small‐cells
Indication that D2D is a component technology that may help realise KPI P3a is provided by the
simulation and analysis of network controlled D2D which indicate potential facilitate support for
very large numbers of devices. Also, the results achieved by the use of different functional splits (HLS
PDCP/RLC, LLS MAC/PHY, LLS UPPER PHI), allowing efficient use of available transport bandwidth,
increases the potential the KPI P3b may be met when deploying these technologies. Additionally,
common transport carrying fronthaul and backhaul with TSN Ethernet fronthaul supports
deployment of femto‐cell, that is small‐cells P3b
P4: “Create a secure, reliable and dependent internet with “zero perceived” downtime…”
with proxy metrics P4a of provide a 1‐10ms latency over the air interface, P4b reach
reliability > 99% and, P4c reach availability > 99%
Support for KPI P4a related to latency is provided by the D2D simulation and analysis that optimises
the set‐up procedures and offers the possibility of direct communication not intermediated by the
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 70 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
network. And, additionally, support for this KPI is provided in the following testbeds; the 60GHz
wireless testbed, which provides latency in the range 250µs ‐ 1200µs dependent on throughput, the
vRAN PDCP‐RLC testbed that shows that existing installed fibre has potential to meet the less
stringent 5G use‐cases regarding latency, and finally, extrapolation based on results of the software‐
based radio with MAC/PHY split, indicate that the system also has potential to meet the KPI.
Additionally, the use of Ethernet based fronthaul also facilitates path redundancy that contributes to
availability P4c.
In summary, the Showcase demonstrator and its component testbeds, and the testbeds
demonstrated separately or examined through simulation, has made many contributions that are
relevant to the 5G KPIs defined by the 5GPPP [1].
5 ConclusionOne of the major highlights was the Showcase demonstration itself, which contributes to the 5G
business KPI B3, with respect to proof of concept demonstration of 5G technologies, and to 5G
societal KPI S3, with respect to maturity of potential 5G component technologies and impact on total
cost of ownership.
The 5G performance KPIs are very demanding, but in several cases the testbeds directly met them
them, for example, demonstration of a 2.2Gbit/s application rate with latency in the sub‐millisecond
range by the 60GHz wireless against a requirement of 1‐10GHz peak user rate and 1‐10ms latency
and. Additionally, investigation showed that the installed fibre base has the potential to meet
latency constraints for some of the less demanding 5G use‐cases, with 5‐10ms versus 1‐10ms
latency. Additionally, the role that other technologies may play in achieving these KPIs was
demonstrated, for example, the HLS and LLS that prevent fronthaul becoming a bottleneck for future
5G and C‐RAN/vRAN deployment, the use of D2D that may assist in achieving the reliability and
accessibility KPIs and energy saving, and the cloud‐assisted applications that facilitate energy saving
for the mobile device.
Other of the societal and performance KPIs are qualitative in nature, however, significant
contribution was shown in the potential to stimulate new services S4, and reducing average service
creation time with the deployment of virtualisation and extensive use of opensource software.
Thus, we can conclude that the iCIRRUS project, and particularly the Showcase demonstrator and its
component testbeds, made significant contributions to understanding and validation of the 5G
business, societal and performance KPIs as set out in [1].
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 71 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
6 References
[1] D. Kennedy, “D2.6 ICT ‐ Final report on programme progress and KPIs,” 5G PPP H2020‐ICT‐2014‐
2, 2017.
[2] P. Chanclou, Z. Tayq, P. Turnbull, V. Jungnickel, L. Fernandez‐del‐Rosal, P. Assimakopoulos, N.
Gomes, K. Yuan, H. Zhu, M. Parker, S. Walker, C. Magurawalage, K. Yang, H. Thomas, M.
Castaño‐Torres, I. Campos Rivera and F. Jesus‐Hidalgo‐Gonzalez, “D2.1 iCIRRUS – Intelligent C‐
RAN Architecture,” Horizon 2020 iCIRRUS Project, 2015.
[3] P. Asimakopoulos, N. Gomes, H. Zhu, L. Rosal‐del‐Fernandez, M. Hinrichs, P. Ritoša, D. Münch, J.
P. Elbers, C. Magurawalage, K. Wang, S. Delaitre, M. Castaño‐Torres and S. Blasco, “D5.2
iCIRRUS Tools, scenarios and results for testing analysis methods for validation test,” Horizon
2020 iCIRRUS Project, 2017.
[4] H. Thomas, L. Fernandez‐del‐Rosal, P. Asimakopoulos, Y. Kai, K. Wang, P. Chanclou and D.
Muench, “D3.3 SLA and SON Concept for iCIRRUS,” Horizon 2020 iCIRRUS Project, 2016.
[5] K. Habel, C. Kottke, M. Hinrichs, L. F. d. Rosal, D. Muench, N. Eiselt, P. Assimakopoulos, N.
Gomes, H. Thomas, M. parker, F. Ngobigha, G. Koczian, T. Quinlan, S. walker and P. Chanclou,
“D3.4 iCIRRUS Updated low‐cost, energy‐efficient fronthaul,” Horizon 2020 iCIRRUS Project,
2017.
[6] P. Ritoša, S. Delaitre, G. Flores, D. Capitán, P. Chanclou, R. Sušnik, H. Zhu, M. Polyvios, J. Zou, P.
Elbers and H. Thomas, “D2.3 iCIRRUS Business case analysis,” Horizon 2020 iCIRRUS Project,
2017.
[7] P. Kangru, “ZSM (18) 000037 RAN SON ZSM Input iCIRRUS,” in ETSI ZSM (18), 2018.
[8] W. Featherstone, D. Padfield, C. Murphy, H. Thomas, P. Claridge and M. Wang, “Enhancing
network topology information for a self‐organizing network”. Patent US201514929923
20151102, 10 05 2017.
[9] J. Haver, H. Thomas and J. Govert, “Packet colouring for data stream monitoring”. Patent Not
yet published, 8 9 2017.
[10] D. Padfield, C. Murphy and H. Thomas, “Techniques for providing fronthaul data awareness”.
Patent US201562147827P, 15 4 2015.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 72 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
[11] P. Turnbull, “Method and apparatus for mitigation of packet delay variation”. USA Patent
US201414562062 20141205, 11 06 2015.
[12] Wikipedia, “Unified communications,” Wikipedia, 17 12 2018. [Online]. Available:
https://en.wikipedia.org/wiki/Unified_communications#cite_note‐1. [Accessed 17 01 2018].
[13] K. Habel, P. Assimakopoulos, N. Gomes, I. Laurinavičius, H. Zhu, M. Hinrichs, S. Weide, J. Zou
and J. Elbers, “D5.3 Validation test setup and execution report,” Horizon 2020 iCIRRUS Project,
2017.
[14] O. Sahin, P. Pietraski, A. Demir, W. Lawton, A. Roy, M. Fazili, K. Pan, M. Abou‐El‐Seoud and K.
Wanuga, “High resolution angle of arrival estimation and dynamic beam nulling”. USA Patent
PCT/US2017/040685, 05 07 2017.
[15] M. Georgiades, P. Michael, P. Chanclou, P. T. H. Ritosa, S. Delaitre, N. Gomes, H. Zhu, P.
Asimakopoulos, L. Fernandez‐del‐Rosal and K. Wang, “D2.2 Refined iCIRRUS architecture
definitions and specifications,” Horizon 2020 iCIRRUS Project, 2016.
[16] K. Wang, X. Du, C. Magurawalage, K. Yang, G. Flores, M. Castaño‐Torres, H. Thomas, P. Ritosa, Y.
Kai and H. Zhu, “D4.4 Joint resource management,” Horizon 2020 iCIRRUS Project, 2017.
[17] G. S. Birring, P. Assimakopoulos and N. Gomes, “An Ethernet‐based fronthaul implementation
with MAC/PHY split LTE processing,” in Globecom, Singapore, 2017.
[18] P. Assimakopoulos, G. S. Birring and N. Gomes, “Effects of contention and delay in a switched
Ethernet evolved fronthaul for future cloud‐RAN applications,” in ECOC, Gothenburg, 2017.
[19] PrimeTel, University of Kent, Orange, InterDigital, HHI, University of Essex, ADVA, Telecom
Slovenia, VIAVI, IAF, “D6.4 Final dissemination, communication, standardisation and
exploitation: Plan and report of activities,” Horizon 2020 iCIRRUS project 2018, 2018.
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 73 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
7 ListoffiguresFigure 2‐1 KPIs for fronthaul and RAN joint optimisation ............................................................................ 21
Figure 2‐2 iCIRRUS architecture overview ................................................................................................... 25
Figure 3‐1 Point‐to‐point with SyncE .......................................................................................................... 27
Figure 3‐2: Test setup for Integration of DU, RU, UE and 60GHz .................................................................. 29
Figure 3‐3: comparison of all setups for different frame lengths: 86bytes (a), 512bytes (b), 1024bytes (c),
1518bytes (d), random size 86…1518bytes (e) .................................................................................... 30
Figure 3‐4 Indicative setup for software‐based‐DU radio over Ethernet ...................................................... 32
Figure 3‐5 Command line interface (CLI) of the video server generating RTMP video stream ....................... 38
Figure 3‐6 Monitoring interface of the CPU usage of mobile clone (video proxy) ......................................... 40
Figure 3‐7 Monitoring interface of the network interface traffic of mobile clone (video proxy) ................... 40
Figure 3‐8 Monitoring interface of the mobile phone battery current ......................................................... 43
Figure 3‐9 Monitoring interface of the mobile phone's battery voltage and capacity ................................... 43
Figure 3‐10 Monitoring interface of the mobile clone's CPU usage .............................................................. 43
Figure 3‐11 Monitoring interface of the mobile clone's network traffic ....................................................... 44
Figure 3‐12 Spectral efficiency achieved by network controlled D2D communication with frequency reuse
among L sub‐cells ............................................................................................................................... 45
Figure 3‐13 Latency Comparison between D2D communications and LTE systems ....................................... 46
Figure 3‐14 Experimental setup (top) PtP, (bottom) PON ............................................................................ 47
Figure 4‐1 iCIRRUS testbed setup at Telekom Slovenije ............................................................................... 49
Figure 4‐2 Metro fibre link (in red) used in the iCIRRUS integration demo ................................................... 51
Figure 4‐3 CPRI/Ethernet bridge connected to aggregator without SyncE .................................................... 52
Figure 4‐4 Latency vs. FIFO buffer level (2.45Gbit CPRI) .............................................................................. 53
Figure 4‐5 Latency vs. FIFO buffer level (9.83Gbit CPRI) .............................................................................. 53
Figure 4‐6: 60‐GHz fronthaul system with radio loopback (Ethernet analyser shown as data source and sink)
.......................................................................................................................................................... 55
Figure 4‐7: Throughput for 60GHz fronthaul link for different frame sizes at different layer 1…4 (L1…L2) .... 56
Figure 4‐8: Latency vs. frame size for 60GHz fronthaul link ......................................................................... 56
Figure 4‐9: Jitter vs. frame size for 60GHz fronthaul link ............................................................................. 56
Figure 4‐10: 60GHz fronthaul system with radio loopback and two aggregator nodes ................................. 57
Figure 4‐11: Throughput for 60GHz fronthaul link with aggregator switches for different frame sizes at
different layer 1…4 (L1…L2) ................................................................................................................ 57
Figure 4‐12: Latency for 60GHz fronthaul link with aggregator switches for different frame sizes ................ 57
Figure 4‐13: Jitter (ave) for 60GHz fronthaul link with aggregator switches for different frame sizes ........... 57
Figure 4‐14: Throughput for 60GHz fronthaul link with 3km fibre for different frame sizes at different layer
1…4 (L1…L2) ....................................................................................................................................... 58
Figure 4‐15: Latency for 60GHz fronthaul link with 3km reach for different frame sizes ............................... 58
Figure 4‐16: Jitter for 60GHz fronthaul link with 3km reach for different frame sizes ................................... 59
Figure 4‐17 Set‐up used for validation testing (left) and interface measurement points (right) .................... 60
Figure 4‐18 Fronthaul latency with two Ethernet switches over a 1Gbps Ethernet network, Mean (Left) and
Standard Deviation (Right). The traces include packet latency (“per‐packet”) and subframe latency
(“per‐subframe). The dotted lines are baseline cases while the rest of the traces are with bursty
background traffic. ............................................................................................................................ 61
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 74 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
Figure 4‐19 Mean and standard deviation for the fronthaul latency with two Ethernet switches over a 1Gbps
Ethernet network. The traces include packet latency (“per‐packet”) and subframe latency (“per‐
subframe). The dotted lines are baseline cases while the rest of the traces are with constant packet‐
rate background traffic. ..................................................................................................................... 61
Figure 4‐20 Benchmarking results for the LLS evolved fronthaul. The fronthaul data rate scales with the
application data rate (equivalent to the cell load) while the overhead remains approximately constant
at around 42%. .................................................................................................................................. 64
iCIRRUS Contract No. 644526 1 Jan 2015 – 31 Dec 2017
Page 75 of 75
This project has received funding from the European
Union’s Horizon 2020 research and innovation
programme under grant agreement No 644526
8 Listoftables
Table 2‐1 Comparison of Unified Communication offerings ......................................................................... 17
Table 2‐2 Compilation of Fronthaul KPI ...................................................................................................... 20
Table 2‐3 KPIs for D2D link set up ............................................................................................................... 22
Table 2‐4 KPIs for D2D communication ....................................................................................................... 23
Table 3‐1 Legacy 2.45Gbit CPRI traffic over Ethernet, requirement v performance ...................................... 28
Table 3‐2 Legacy 9.83Gbit CPRI traffic over Ethernet, requirement v performance ...................................... 28
Table 3‐3 lower layer split at the UPPER PHY, requirement v performance.................................................. 31
Table 3‐4 lower layer split at the MAC/PHY, requirement v performance .................................................... 33
Table 3‐5 Component of the video streaming use‐case ............................................................................... 38
Table 3‐6 Measurements for mobile clone .................................................................................................. 39
Table 3‐7 KPIs for mobile clone .................................................................................................................. 40
Table 3‐8 Component of the mobile task offloading use‐case...................................................................... 41
Table 3‐9 KPIs for mobile phone ................................................................................................................. 42
Table 3‐10 KPIs for mobile task off‐loading ................................................................................................. 44
Table 3‐11 D2D Communications performance ........................................................................................... 46
Table 3‐12 HLS split at the PDCP‐RLC layer, requirement v performance ..................................................... 48
Table 4‐1CPRI over Ethernet ‐ MAC payload‐sizes ....................................................................................... 50
Table 4‐2 Legacy 2.45Gbit CPRI traffic over Ethernet, requirement v performance ...................................... 54
Table 4‐3 Legacy 9.83Gbit CPRI traffic over Ethernet, requirement v performance ...................................... 54
Table 4‐4 lower layer split at the UPPER PHY, requirement v performance.................................................. 59
Table 4‐5 Measurement results for the packet and subframe latencies for the baseline cases (no contention)
(x indicated not measured) ................................................................................................................ 62
Table 4‐6 Estimated packet and subframe latencies for the baseline case (no contention) for 10GbE
networking (x indicates not measured) .............................................................................................. 62
Table 4‐7 Mean and standard deviation (STD) of the packet and subframe latencies with the iCIRRUS
aggregation set‐up. The values are extrapolated from the measured results....................................... 62
Table 4‐8 MAC‐PHY LLS, requirement v performance .................................................................................. 63
Table 4‐9 Backhaul (unified transport), requirement v performance ........................................................... 66