Investigating Depth-Fanout Trade-Off in WiMAX Mesh Networks Salim Nahle Luigi Iannone Benoit Donnet Timur Friedman Laboratoire LIP6 – CNRS Université Pierre.

Investigating Depth-Fanout Trade-Off in WiMAX Mesh Networks

Salim NahleLuigi IannoneBenoit Donnet

Timur Friedman

Laboratoire LIP6 – CNRSUniversité Pierre et Marie Curie – Paris 6

First Weird Workshop on WiMAX, Wireless and Mobiliy22-05-2007

2

Overview

IntroductionDepth-fanout trade-offA traffic model for mesh treesSimulationConclusions and future work

3

Introduction

Wireless mesh networks (WMNs) have many advantages.

802.11-based WMNs have been widely studied. Problem: short range. Consequence: dense and suboptimal deployments.

IEEE 802.16 promises to transcend this limitation. It operates in two modes: PMP and MESH

4

Introduction

5

Introduction

6

Introduction

7

Introduction

Dead zones

Long links

Route around obstacles

Multiple shorter hops

What is

the best

number of

hops ?

8

Overview

IntroductionDepth-fanout trade-offA traffic model for mesh treesSimulationConclusions and future work

9

Trade-off: fanout vs depth

3000m

830m

600m 600m

Maximum depth = 2 Maximum depth = 7

10

Hop distance vs. depth

0

0,5

1

1,5

2

2,5

3

1 2 3 4 5 6 7

Tree depth

Dis

tan

ce i

n K

m Depth vs average hop distance

11

Depth vs control overhead

Overhead

0102030405060708090

1 2 3 4 5 6 7

Tree depth

Nu

mb

er o

f co

ntr

ol

tim

e sl

ots

12

Bit rate as a function of Distance

Graph from Betancur et al. NS2 workshop 2006

Distance (m)

Bit

rate

(M

bps)

13

Example

4KmData Rate = 2.2Mbps

2Km 2Km

7.2 Mbps 7.2 MbpsData Rate = 3.6Mbps

DS

DS 1.3Km 1.3Km

11 Mbps 11 Mbps

1.3Km

11 Mbps

DS

Data Rate = 3.6Mbps

2.2 Mbps

14

Overview

IntroductionDepth-fanout trade-offA traffic model for mesh treesSimulationConclusion and future work

15

WiMAX Mesh mode (Background) Mesh mode specifications were integrated into the IEEE 802.16-2004.

They define the control mechanisms and management messages to establish connections in Mesh Network architecture.

16

Traffic model for mesh tree (1/2)We assume a balanced or a quasi-balanced tree.

Parameters:Ca: average number of children SSs per parent node

m: tree depth or number of levels in the WiMAX mesh tree.

17

Traffic model for mesh tree (2/2)Four types of traffic patterns at each SS:

Traffic in the uplink direction towards the Internet via the BS Traffic in the downlink direction from the Internet via the BS Intra-mesh traffic in the uplink direction Intra-mesh traffic in the downlink direction

Note that, within the mesh context, uplink and downlink are defined as the traffic in the direction of the mesh BS and traffic away from the mesh BS repectively.

18

Traffic via BS (1/2)

l

l

l l

l

l l

u u u u

3u 3u

l

l

l l

l

l l

u u u u

3u 3u

7u 7u

BS

7d7d

3d 3d 3d 3d

d d d d d dd d

Traffic per link

Note : l =d+u

3l3d3uA1-A2

lduA2-A3

7l7d7uA0-A1

TotalDownlinkUplink

u: outgoing own traffic

d: Incoming own traffic

l: Average total traffic

19

lC ia *

m

ik

kaCl

im

k

kai lCL

0

*

Average traffic on all the links Ai–Ai-1 :

= (Number of nodes in level Ai)*l + (Average traffic from Ai+1)

= + {(Number of nodes in level Ai+1)*l + (Average traffic from Ai+2)}

=

Average traffic per link:

Traffic via BS (2/2)

20

u13

3

u13

1

u13

11

u

BSCa=2

N=2+22+23=13

Intra-mesh traffic per node = u

u13

1

u13

3

u13

1u

13

1 u13

1u

13

1u

13

1 u13

1

u13

1

u13

1u

13

1

u13

7

u13

1

u13

3

u13

7

u13

1

u

Traffic sent=

Traffic received

Intra-mesh Traffic (1/2)

21

1*

1

1*

11

i

aaia

ii U

N

CCCCu

N

CU

im

i

kai CC

1

Ui+1

Ui

u

u u

u

u u

u

u

u u…

u

Ui+2

1*1

11

1

i

a

ia

UN

CC

C

1*1

11

ia

ia

UN

CC

C

uN

Ci *1

uN

Ci *1

1

1*1

11

1*

i

a

ia

a UN

CC

C

C uN

Ci *1

1

Ui = +

With

Intra-mesh Traffic (2/2)

22

Overview

IntroductionDepth-fanout trade-offA traffic model for mesh treesSimulationConclusions and future work

23

Simulation

Objective: Study the impact of the tree depth m on the aggregate throughput

capacity.

Simulation setup: Number of nodes = 49

Distributed uniformly in a 7*7 grid topology

m varies between 1 and 7

Traffic Requests are sent as just described

24

Throughput via the BS

0

1

2

3

4

5

6

7

8

1 2 3 4 5 6 7

Tree Depth

Agg

rega

te th

roug

hput

in M

bps

25

Intra-mesh throughput

0

0,5

1

1,5

2

2,5

3

3,5

4

4,5

5

1 2 3 4 5 6 7

Tree depth

Ag

gre

gat

e th

rou

gh

pu

t in

Mb

ps

26

Overview

IntroductionDepth-fanout trade-offA traffic model for mesh treesSimulationConclusions and future work

27

Conclusions

Increasing the depth may increase the throughput even without allowing for concurrent transmissions

Only long hops must be split.

Recent extension shows better results

28

Future work

Different traffic models

Allowing for concurrent transmissions (future work).

Investigate distributed scheduling capacity.

Optimizing the number of time slots used by each scheme is another perspective.

29

Questions

Thank you