EU FP7 Project iJOIN
Transcript of EU FP7 Project iJOIN
http://www.ict-ijoin.eu/ @ict_ijoin
EU FP7 Project iJOIN
Cloudification of the 5G Radio Access Network
ITG Zukunft der Netze 2014
Sept. 26th 2014 Andreas Maeder, Peter Rost (NEC Laboratories Europe)
Contact: {andreas.maeder, peter.rost}@neclab.eu
iJOIN: Interworking and JOINt Design of an Open Access and Backhaul Network Architecture for Small Cells based on Cloud Networks
http://www.ict-ijoin.eu/ @ict_ijoin (( 2 ))
Laws of growth: drivers of cloudification
GSM GPRS
UMTS EDGE
HSDPA
WiMAX HSPA+
LTE LTE-A
1
10
100
1.000
10.000
100.000
1.000.000
1990 1995 2000 2005 2010 2015 2020
Cellular peak data rates over time
Dat
a R
ates
[kBi
t/s]
8088 80486
Pentium Pro
Pentium 4
AMD K8 Core 2 Duo Core i7 Opteron
Sparc T3
Intel Xeon Phi
0,1
1
10
100
1000
10000
1970 1980 1990 2000 2010 2020
Tran
sist
or C
ount
(10^
6)
Transistor Density
0,1
1
10
100
1000
1985 1990 1995 2000 2005 2010 2015
Sotra
ge D
ensi
ty (G
Bit/i
n²)
Storage Area Density Storage
Commu- nication
Processing
http://www.ict-ijoin.eu/ @ict_ijoin (( 3 ))
• SaaS • PaaS • IaaS
• On-demand • Broad access • Pooling • Elasticity • Measured
• C-RAN • RAN-Sharing • SDN • SDR
How the “Cloud” changes the picture …
SQL ISP
• NFV • vEPC • SDN
Core Network HetNet RAN Heterogeneous Backhaul
60GHz
Relay/Microwave
Copper
Optical
Mobile Network
Virtualization / “Cloudification”
The communication realm The IT world
Blurring borders between Communication and Information Technology
http://www.ict-ijoin.eu/ @ict_ijoin (( 4 ))
Key enablers to satisfy data demands
0
200
400
600
800
1000
1200
1400
1600
MoreSpectrum
FrequencyDivision
Modulationand Coding
SpectrumReuse
25 5 5
1600
Centralised Processing • C-RAN handles inter-cell interference, allows for higher
utilisation and to avoid peak-provisioning • Up to 50% energy-saving • 20%-50% OPEX reduction, 15% CAPEX reduction • Requires high capacity and low delay backhaul Centralisation is an option to implement the network but
requires more flexibility than today
Small Cells • 50% Total cost of ownership (TCO) savings • Four-fold increase in density until 2014 • Worth about 6.1bln USD until 2014 Small-cells are the option to handle higher rates and to
improve energy/cost-efficiency Fact
or o
f im
prov
emen
t
http://www.ict-ijoin.eu/ @ict_ijoin (( 5 ))
Key Concepts
Flexible centralisation through RANaaS (RAN-as-a-Service) Push RAN functionality into cloud-like platforms Simplified RAN management and flexible small-
cell solutions Reduce complexity & cost through elastic &
flexible function assignment Higher energy-efficiency through computational
diversity and higher utilisation
Joint design and optimisation of RAN and backhaul Interworking of access and backhaul network Optimise for flexible centralisation Optimise backhaul for small cells Consider heterogeneous backhaul network
(fibre and wireless) Relax backhaul requirements through dynamic
provisioning (“on-demand”)
Cloud Platform Core
Network
eNB
iSC iSC
iSC
iSC
iSC
eNB
RANaaS
Join
t Opt
imis
atio
n of
RA
N a
nd B
ackh
aul
iTN
RANaaS
LTE Air I/F mmWave Fibre Link
Very
den
se s
mal
l ce
ll ne
twor
k H
eter
ogen
eous
ba
ckha
ul la
yer
Flex
ible
RAN
func
t. as
sign
. to
clou
d
iSC: iJOIN Small Cell
http://www.ict-ijoin.eu/ @ict_ijoin (( 6 ))
Key Concepts: flexible functional split
RF
PHY
MAC
RRM
Netw. Mgmt.
Adm./Cong. Control
“Conventional” implementation of LTE
RF
PHY
MAC
RRM
Netw. Mgmt.
Adm./Cong. Control
C-RAN Implementation (BB-pooling)
Exe
cute
d at
BS
Centrally executed
Cen
trally
ex
ecut
ed
Executed at R
RH
Flexible Functional Split
Example: Partly centralised (inter-cell) RRM
Example: Joint Decoding
http://www.ict-ijoin.eu/ @ict_ijoin (( 7 ))
Key concepts: exploit centralization gains
Computational diversity Exploitation of temporal and spatial traffic fluctuations Efficiently use available resources, scale resource according to needs
(resource pooling, elasticity)
RAN
RANaaS
RANaaS Interface RANaaS Hypervisor
http://www.ict-ijoin.eu/ @ict_ijoin (( 8 ))
Key concepts: service and function agility
Localized optimization Optimization based on purpose, deployment, … Using software implementation rather than configuration (SON) Flexible software assignment over time and space
RAN
RANaaS
RANaaS Interface RANaaS Hypervisor
http://www.ict-ijoin.eu/ @ict_ijoin (( 9 ))
Challenges
Feasible and beneficial functional splits
Architecture
Data center placement
Utilization efficiency
Resource management of virtualized environment
Cost efficiency
http://www.ict-ijoin.eu/ @ict_ijoin (( 10 ))
Challenges: Where to split?
MA
C
HARQ Rtx
Multiplexing
Priority Handling
PH
Y-
FEC
Modulation
Etc.
PDCP
RLC
Buffering
Segmentation/ARQ
GTP
Scheduling
RRC
- More bandwidth - Lower Latencies
+ More gain?
Constraints and requirements: Backhaul characteristics
(latency, bandwidth) 3GPP requirements (latency,
bandwidth, protocol stack) Computational resources
Benefits:
Multiplexing gains (computational resources, user traffic)
Coordination gains (interference mitigation, cancellation, load balancing, ...)
..are an operator-defined optimization target!
http://www.ict-ijoin.eu/ @ict_ijoin (( 11 ))
RLC
PDCP
MAC
PHY
RRC
Functional split options
Minimal centralization requirement to achieve centralization gains: Main scenarios:
• Centralized data processing (PHY+higher layers) • Centralized resource allocation • Centralized connection control U-Plane
C-Plane
Centralized Resource Allocation
RANaaS SC
RLC
PDCP
MAC
PHY
Centralized Connection Control
LRRC GRRC
Scheduling
RANaaS: Central RRM iSC: Local RRM
RANaaS: global RRC iSC: Local RRC
RANaaS SC
CRRM Measurements, etc
LRRM
http://www.ict-ijoin.eu/ @ict_ijoin (( 12 ))
LTE Bandwidth demands
100% 1.1%
LTE UL 20Mhz 50% RB utilization 2 Rx Antennas SE 3 bit/s/Hz
user diversity gains
http://www.ict-ijoin.eu/ @ict_ijoin (( 13 ))
Challenges: network architecture
J2
http://www.ict-ijoin.eu/ @ict_ijoin (( 14 ))
Challenges: Data Processing Capabilities
Computational complexity depends on Functional split Number of connected base stations Number of connected devices Channel and traffic conditions
Strongly non-linear complexity in SNR
http://www.ict-ijoin.eu/ @ict_ijoin (( 15 ))
Computational complexity
System-level results for per-cell CC Correlation between #UEs and CC? Impact of scheduling, resource allocation?
3.6 3.65 3.7 3.75 3.8 3.85 3.9 3.95 4 4.05x 104
0
2
x 104
subframe index
tota
l com
p. c
ompl
exity
(bits
⋅ it)
3.6 3.65 3.7 3.75 3.8 3.85 3.9 3.95 4 4.05x 104
0
5
num
ber o
f UE
s pe
r cel
l
Comp. compl.#UEs
http://www.ict-ijoin.eu/ @ict_ijoin (( 16 ))
Challenges: Utilization Efficiency
0 5 10 15 20 25 30 35 40 45 5020
40
60
80
100
120
140
Number of cells N
Util
izat
ion
[%]
log10(ε) = -2log10(ε) = -1.75log10(ε) = -1.5log10(ε) = -1.25log10(ε) = -1
Utilization in current systems below 40% Utilization in centralized systems increases… but how much? Computational outage vs. channel outage should be same order
90+% Utilization can be achieved with centralized RAN
http://www.ict-ijoin.eu/ @ict_ijoin (( 17 ))
Challenges: Data Center Placement
Access Aggregation Metro Core
Internet
small cell
RANaaS
eNB
P/S-GW
iTN
Wired link
p2p wireless
P2mp wireless
C1
C3
C4
C5
C7
C6
C2
http://www.ict-ijoin.eu/ @ict_ijoin (( 18 ))
Challenges: cost-efficiency
0 2 4 6 8 101.5
2
2.5
Data center density [per sq-km]
CA
PEX
[MEU
R p
er s
q-km
]
DRANCRAN, LTE performanceCRAN, 0.4dB offsetCRAN, 0.9dB offset
BH Data centers
http://www.ict-ijoin.eu/ @ict_ijoin (( 19 ))
Cloudification decision variables
Functional split /
deployment
Constraints: backhaul, cell/user density,
protocols
Computational resource
availability
Operator optimization
target
Cost
http://www.ict-ijoin.eu/ @ict_ijoin (( 20 ))
Conclusions
New paradigms Operate RAN based on commodity hardware Resilience to computational and communication imperfections Allow for similar reliability on computation side as on communication side
New insights
Cloud RAN is cost-efficient when planned and deployed appropriately Computational demands are main drivers; depends on RAN resource allocation Many more results available on www.ict-ijoin.eu, e.g. feasible splits, BH
dimensioning, data center placement, CoMP, commodity HW implementation, …
Outlook: RAN cloudification will be integral aspect of 5G Standardization bodies will take up (e.g. small cell forum) IT companies will continue to push into telco space
http://www.ict-ijoin.eu/ @ict_ijoin (( 21 ))
Thank you for your attention!
http://www.ict-ijoin.eu/ @ict_ijoin (( 22 ))
Backhaul Technology Characteristics
BH technology Latency (one-way) Throughput
Ideal fiber access 2.5 µs 10 Gbps
Fiber Access 1 10 ms – 30 ms 10 Mbps – 10 Gbps
Fiber Access 2 5 ms – 10 ms 100 Mbps – 1 Gbps
DSL Access 15 ms – 60 ms 10 Mbps – 100Mbps
Cable 25 ms – 35 ms 10 Mbps – 100 Mbps
Sub -6 GHz Wireless 5 ms – 35 ms 50 Mbps – 100Mbps
Microwave < 1 ms 100 Mbps – 1 Gbps
mmW radio < 1 ms 500 Mbps – 2 Gbps
Diverse characteristics Significant impact