The Meraka Indoor wireless mesh test bedThe Meraka Indoor wireless mesh test bedA new multi hop routing benchmarking toolA new multi hop routing benchmarking tool

David Johnson

Senior Researcher

Wireless Africa Programme

Meraka

CSIR

450 million rural people in Africa areisolated from the rest of the world

Connect them all by 2017

Sound crazy?

How do these sound?

7 trillion wireless devices serving 7 billion people by 2017 …Wireless World Research Forum

Affordable mobile communications for half the world's population by 2015 …Nokia

Half the world have Internet access by 2015 …AMD 50x15

Creates cognitive dissonance

Here’s how it could be done

Turn the “first mile” into an open bazaar of community networks

As easy to setup as a TV

Telcos just provide backhaul

Income stays in the communities

Modernized without being urbanized

Ubuntu = community networks = mesh networks

• Plenty of research problems to address– Routing protocols and routing metrics– Gateway selection for multiple gateways– Power efficiency– Auto channel allocation– Distributed services– Multimedia over mesh– Clustering optimization– Smart antennas for mesh– Business models for village entrepreneurs– Security over mesh– Vulnerability of first time exposure to the internet– Building technical skills in developing regions– Logistics

Why build a test bed?

Mathematical models good enough for rules of thumb but no tractable solution for a real network

))log(

()(nn

WnWORST )()(

n

WnBEST

Simulation models good for early protocol development and testing but still far from reality

Real outdoor wireless = reality but difficult to build and manage

Indoor test beds provide controlled environment and expose theoretical simplifications

Construction of the meraka wireless grid

Construction of the meraka wireless grid

Construction of the meraka wireless grid

Construction of the meraka wireless grid

Construction of the meraka wireless grid

Challenges when building experiments

Challenges when building experiments

2 4 6

246

2 4 6

246

2 4 6

246

linux-valuesT3a-1Mbps-ch8_Tx19dBm.dat2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

2 4 6

246

rssi

0

5

10

15

20

25

Challenges when building experiments

• Which combination of mode/rate/txpower to use• Experiments are time consuming ... 52 hours to test 4

routing protocols using all combinations of 49 nodes in the grid with 20 second test time

Electromagnetic modelling of the grid

Done with WIPL-D modelling software

Electromagnetic modelling of the grid

0 0.8 1.6 2.4 3.2 4 4.8 5.60

5

10

15

20

25

30

RS

SI

distance between nodes, m

meas: 11bDS1Mbps-5ch8model: ant1_box_7x7model: wire_model_7x7

Results – 49 node multi hop chain

Results – 49 node multi hop chain

)()(98.0_ n

WnLABMERAKA

Results – 49 node multi hop chain

• After 4 hops – Gupta best = 50%– Gupta worst = 42.47%– Gupta indoor measurements = 9.74%– Meraka lab = 25.7%

Choosing routing protocols to benchmark

Results – hop count distribution

Results –Routing traffic overhead

Results –Routing traffic overhead

Results – throughput, packet loss, delayString of 7 nodes

Forward HOP count Route changes Packet loss Delay Delay(stddev) TP No link

AODV 1.33 0.43 11.19 37.24 116.64 2723.36 1DYMO 1.52 0 9.52 3.65 2.37 2907.67 0OLSR_ETX 1.43 0.1 8.57 27.56 101.91 2730.69 0OLSR_RFC 1.67 0.76 2.14 5.35 5.35 2923.64 0

Results – throughput, packet loss, delayFull 7x7 grid

Forward HOP count Route changes Packet loss Delay Delay(stddev) TP No link

AODV 1.36 0.53 71.22 117.87 317.35 773.33 1425DYMO 2.2 0.11 32.81 64.72 150.2 1165.66 413OLSR_ETX 1.84 0.25 24.05 68.84 247.78 1187.57 453OLSR_RFC 2.28 2.34 22.22 67.44 132.49 1330.05 381

Comparison of throughput to baseline

Conclusions

• Proved that multi hop wireless networks possible on scaled test beds (up to 5 hops were achieved in a space of 7Mx7M)

• Complexity in the grid is high and produces a worst case scenario for routing algorithms.

• The AODV protocol showed the weakest performance but had the lowest routing overhead.

• DYMO showed good results for its low routing overhead with the least amount of delay for the full 7x7

• The RFC version of OLSR had the best overall performance the full 7x7 grid.

• OLSR with the ETX extension performed better in medium size networks of about 21 nodes.

Future improvements to the mesh lab

• Compare results to NS2 simulations

• Add a lab sandbox• Add a lab scheduler to make it

simple for multiple interested parties to run experiments

• Build a measurement framework to simplify setting up experiments

• Test virtual mobility and install a roaming robot

• Experiment with parameters beyond the RFC values for routing protocols

• Add support for power measurements

• Add support for control of smart antennas

For further information

http://wirelessafrica.meraka.org.za