EU FP7 Project iJOIN - TUM Challenge: computational complexity Main contributor to computational...

http://www.ict-ijoin.eu/ @ict_ijoin

EU FP7 Project iJOIN

Functional Split Options for Centralized RAN with Imperfect

Backhaul

46. Treffen der VDE/ITG-Fachgruppe 5.2.4

Mobilität in IP-basierten Netzen Dec. 4th 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/



http://www.ict-ijoin.eu/ @ict_ijoin (( 2 ))

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 Facto

r o

f im

pro

ve

me

nt





• 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





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 O

ptim

isation

of

RA

N a

nd

Backhaul

iTN

RANaaS

LTE Air I/F

mmWave

Fibre Link

Ve

ry d

en

se s

ma

ll

ce

ll n

etw

ork

He

tero

ge

ne

ou

s

backh

au

l la

ye

r

Fle

xib

le R

AN

fu

nct.

assig

n. to

clo

ud

iSC: iJOIN

Small Cell





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)

Execute

d a

t B

S

Centra

lly e

xecute

d C

en

trally

exe

cu

ted

Exe

cu

ted

at R

RH

Flexible Functional Split

Example: Partly centralised

(inter-cell) RRM

Example: Joint

Decoding





Key concepts: exploit centralization gains

Coordination gains: e.g. ICIC, joint precoding, joint decoding, … 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





Key concepts: service and function agility

Localized optimization Optimization based on purpose, deployment, …

Using software implementation rather than configuration (SON)

Fast deployment of tailored software depending on location and use case

RAN

RANaaS

RANaaS Interface RANaaS Hypervisor





Challenges

Functional split of RAN protocol stack

Computational resource provisioning

“cloudified” RAN architecture

Utilization efficiency

Cloud platform latency/U-plane capabilities





Challenge: Where to split?

- higher requirements

(more bandwidth, lower

latencies, higher comp.

effort)

+ gains from

coordination, diversity,

joint processing

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! RF processing

A/D conversion / pre-processing

Lower PHY layer (incl. FFT)

Upper PHY layer (incl. FEC)

MAC layer (incl. MUX, HARQ)

RLC Layer (RB buffers, ARQ)

PDCP Layer (ciphering)

RRC/C-Plane

Sch

edu

ling

Radio Access Point

EPC





Functional Split Options

RF processing

A/D conversion / pre-processing

Lower PHY layer (incl. FFT)

Upper PHY layer (incl. FEC)

MAC layer (incl. MUX, HARQ)

RLC Layer (RB buffers, ARQ)

PDCP Layer (ciphering)

RRC/C-Plane

Sch

edu

ling

Radio Access Point

EPC

Split A

Split B

Split D

Split C

Split option

Lowest layer centralized

RTT requirements

Bandwidth requirements

Centralization scheme

Main centralization gains

D PDCP > 50ms U-Plane + C-Plane overhead

Centralized connection control

Load balancing, energy efficiency

C MAC <3 ms (HARQ sched.)

U-Plane + C-Plane overhead

Coordinated resource allocation

Interference mitigation, cooperative schemes

B FEC ~ 1ms (HARQ + FEC processing time)

U-plane + redundancy (~ +30%)

Coordinated resource allocation, central FEC

Computational diversity

A FFT/IFFT 5 µs (CIPRI)

Quantized symbols (several Gbps)

Full PHY centraliz-ation

Centralized precoding, joint decoding

For more information see iJOIN deliverables D2.2, D3.2





Bandwidth requirements (DL)

user diversity gains

20MHz BW

2x2 MIMO





Latency requirements: HARQ

HA

RQ

RTT

tim

er

sub

fram

e

ind

ex

UE iSC RANaaS

BH latency

Air int.

RANaaS processing

+ frame building

Air int.

BH latency

UL DL

UL data

UL data

Latency breakdown

0

1

2

3

4

5

6

7

8

9

0

1

2

3

4

5

6

7

8

0

1

2

3

4

5

6

7

8

0

1

2

3

4

5

6

7

8

UL DL

9

0

1

2

3

4

5

6

7

1

2

3

4

5

6

7

8

9

UL DL

ACK/NACK

AC

K/N

AC

K e

xpec

ted

in S

F n

+4

ACK/NACK

“Latency bottleneck”, applies

to all splits with centralized MAC





Mapping functional splits to backhaul types

Fiber (CWDM/dark)

mmWave

μ-Wave

Sub-6 GHz

PON

xDSL

Metro-optical

Split A

Split B

Split C

Split D

10 1

10 2

10 3

10 4

10 5

10 -4

10 -3

10 -2

10 -1

10 0

Throughput [Mbps]

La

ten

cy (

pe

r h

op

, R

TT

)[s

]





Challenge: computational complexity

Main contributor to computational complexity is the turbo decoding (UL)

Strongly non-linear complexity in SNR, depends on modulation and coding scheme

Computational outage: TB cannot be decoded within time budget (<3 ms)

P. Rost, S. Talarico, M. Valenti, The Complexity-Rate Tradeoff of

Centralized Radio Access Networks, submitted to IEEE TWC





Computational complexity: cell level

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 4.05

x 104

0

2

x 104

subframe index

tota

l co

mp

. co

mp

lexity (

bits

it)

3.6 3.65 3.7 3.75 3.8 3.85 3.9 3.95 4 4.05

x 104

0

5

nu

mb

er

of U

Es p

er

ce

ll

Comp. compl.

#UEs





Joint RAN/cloud resource provisioning

0 100 200 300 400 500 600116

118

120

122

124

126

128

130

132

134

Frame number

Com

puta

tional dem

and (

CP

Us)

Provided CPUs

Occupied CPUs

Comp.-aware LA

Overall throughput loss: 0.04%

Assumptions: 456 cells/142 base stations (macro+small cell), 1200 users

96 GFlops per CPU

1. Estimate outage capacity based on

per-cell CC outage rate (e.g. 1%)

2. Provide resources (e.g. CPUs)

3. Adjust LA when necessary





Conclusions

New paradigms RANaaS: Apply Cloud principles to RAN operation

Flexible functional split: adapt level of centralization to backhaul capabilities

Increase utilization efficiency by joint cloud/RAN resource management

New insights Feasible functional splits of the LTE protocol stack

“Cloud-aware” link adaptation: first step towards RAN operation on commodity

hardware

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

IT companies will continue to push into telco space

Technical challenges remain: RAN-capable cloud platforms








Thank you for your attention!




EU FP7 Project iJOIN - TUM Challenge: computational complexity Main contributor to computational...

Documents

Transcript of EU FP7 Project iJOIN - TUM Challenge: computational complexity Main contributor to computational...