The role of virtualisation in the dense wireless networks of the future Sokol Kosta CINI.

The role of virtualisation in the dense wireless networks of the future

Sokol KostaCINI

Distributed Computing, Storage, and Radio Resource Allocation over Cooperative

Femtocells

3

Increasing demands in the wireless world

Want it all!Want it here!Want it now!

Can we enrich users experience with today’s terminals?

Well, it is all about apps…

Backhaul and wireless

4

Current elements in modernization of applications

According to CISCO

1. Make them accessible over the Internet on any device anywhere

2. Enhance performance, resilience and throughput

3. Use cloud computing for faster time-to-market, continuous development and change

4. Leverage open source components and open APIs

5. Develop and execute on infrastructure shared by multiple applications and users

TROPIC encompasses these targets and goes beyond…

5

Overall TROPIC objectives

TROPIC redefines architectures able to…

…virtualise/distribute applications close to user at empowered smallcell base stations

…enhance physical layer performance

Bringing computational power closer to users will entail

1. improving user experience, 2. prolong UE battery lifetime, and 3. potential revenue stream for operators

6

Small cell manager

The TROPIC scenario

7

Offloading to small cell eNB: pros and cons

Running apps in empowered small cell eNB instead of

external cloud

Running apps in empowered small cell eNB instead of UE

Small CellManager

+ Reduced latency

+ Reduce usage of backhaul

- Management of virtual machines

+ Computation speed

+ Reduced battery consumption

- Management of parallelization

- Increased PHY utilization Small cell manager (SCM) is needed to allocate computational resources

Technical scenarios

SCM serving single cell

SCM serving multiple cells

Coexistence of multiple SCM and cells

9

Small Cell Manager (SCM) provides offloading support to the UE, including computation/storage/radio resource management Best network architecture may differ among Mobile Network Operators (MNO), and may depend on the scenario (corporate, public, residential)

Network architecture needs to be enhanced

10

Joint management of radio and computational resources

0,1

maxi iUL P P DL

iL t s t

TLatency budget:

0,1, ,P UL DLs s s Bits processed, communicated

from/to UE,UL DLt t Time for UL and DL wireless tx

0,1

2 1 2 1UL DL

UL UL DL DL

i

UL DL

s s

W t W t

T UL P Pi DL Screeni

E t s t p LSNR SNR

EEnergy budget:

Parallel processing

Energy spent in UL

Energy spent in DL

Both convex problems with unique solution

, , , ,min

s.t. UL DL Pi UL DLt t s s s

E

L T

, , , ,min

s.t. UL DL Pi UL DLt t s s s

L

L

E

It is possible to plug an abstration of the PHY layer, for MIMO or even MU-MIMO channels

11

- Given latency and battery contraints, we can select a different operation point

- Each combination of application, processing architecture, and radio tx scheme will generate a different curve

5 10 15 20 25 30 350.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5Energy spent only by the UE

En

erg

y

Maximum Latency (s)

Optimization of the energy spent by the UE only

Optimization of the total energy

SNRDL=SNRUL = 20 dB

Energy vs. Latency Tradeoff

Example for virus scan application…

12

… in a multiuser scenario

UL Remotecomp.

resources

DL

.

.

.

.

.

.

.

.

.

UL schedulerComputation

schedulerDL scheduler

Assumptions:• Multiple users generate offloading petitions at a certain rate: • The offloading decision has already been taken for each UE involved• A certain probability of latency performance can be guaranteed

If multiple users are served by a single FAP, combine energy-latency resource allocation with scheduling policies.

Benefits and potential hurdles

13

Potential hurdles…

• Additional investment by MNO• Best network architecture may differ among MNO, and

may depend on the scenario (corporate, public, residential)

• Improving user experience,• Prolong UE battery lifetime, and• Potential revenue stream for operators

Benefits

Connectivity management for eneRgy Optimised Wireless Dense networks

Goal of CROWD

To enable sustainable deployment of very dense and heterogeneous wireless networks

sustainable = cost effective + energy efficientvery dense = 1000x compared to currentdensity (in users/sqm vs. users/BS)heterogeneous =

+ diverse range (macro vs. pico vs. femto)+ diverse technologies (LTE vs. WiFi)+ diverse deployments (planned vs. unplanned)+ diverse backhaul types (optical vs. wireless)

Research challenges

Very high density + heterogeneity =1. Interference in the radio access network2. Poor efficiency in the backhaul 3. High signalling for traffic management

InterferenceIssues with increasing density to grow capacity:• OPEX: limiting factor, no automated tool for HetNets• CAPEX: base stations are complex & expensive• Capacity does not scale with base station density!!

Some inter-cell cooperation tools are available• ABSF (LTE), OBSS (802.11) • Control tools needed CROWD Control architecture

Increasing number of eNBs provides an increasing of capacity up to a limit.

Indiscriminate increaseof network densityis not a viable solution

Backhaul

Increased capacity demand and density has to be sustained by the backhaul network, not only by the radio access network→ pitfall: costly overprovisioning of the backhaul network• CAPEX: expensive high capacity backhaul equipment• OPEX: increased cost and energy waste

Dynamic provisioning of backhaul resources is the key to an efficientbackhaul operation in very dense networks

Signalling

1. Technology-specific issues, e.g.:– 70% of air time (WLAN) occupied by management frames– Most of air time is taken by scanning processes

2. Mobility-related signaling critical:– Very frequent handovers, lots of subscribers– Reduce signaling over the air, Proxy approaches

CROWD Architecture

Logical view

Business view

The role of virtualisation in the dense wireless networks of the future

Sokol KostaCINI