Adaptive Manet Routing:
A Case Study

Thomas Kunz

Systems and Computer Engineering

Carleton University

Adaptive Routing for MANET

Adaptive Routing:

Mobile node adjusts its routing behavior based on its environment to improve routing performance

Adjustment is on a per-node basis, not global

Levels/Timescales of Possible Adjustments:

Short term: change its routing parameters based on environment

Mid-term: select stable route with enough bandwidth during Route Discovery

Long-term: select adequate routing protocol based on application etc

Adaptive: What do I mean?

“Environment” means many things

Radio environment: could impact route stability and throughput

Traffic pattern: which flows, between which nodes

Mobility: topology changes, formation/destruction of links

First step: adapt to mobility at short time-scales

Identify mobility metric that

Captures the mobility-driven impact of routing protocol performance

Can easily be measured by a node, no extra hardware required

Is true across a number of mobility scenarios (different patterns/mobility models) and independent of mobility model parameters (RWP: Max/min speed, pause time, …)

Based on the metric, modify protocol to allow individual nodes to adapt their protocol behavior

First Simulations: Mobility Metrics

Simulation Results: PDR vs. Mobility

Simulation Results: Link Duration

Simulation Results: Link Duration as Performance Predictor

Simulation Results: Link Breaks

Mobility Metrics: Individual Nodes

Conclusions

OLSR Modifications

OLSR has 4 control parameters

Hello Interval

TC Interval

MPR Coverage

TC Redundancy

Idea is to set parameters “appropriately” for network

They then apply to all nodes and for the whole time

Our modifications:

Monitor link breaks, define two thresholds (upper, lower)

Choose appropriate Hello Interval

more observed link breaks => higher mobility scenario => faster Hello

Add states to OLSR nodes that govern processing of Hello messages/selection of MPRs

Default OLSR, Fast OLSR, Fast Response

OLSR Node States

Simulation Setup

Implemented Adaptive OLSR in NS2 and evaluated performance through experiments

UM-OLSR version 0.8.8 for NS2 version 2.29

The simulation area is 1000x1000m

80 mobile nodes

Default IEEE 802.11 configuration: 250 m transmission range, 11 Mbps data rate

Traffic: 25 data sources, 4 packets/s, 64 byte packet size (CBR)

Simulation time is 900 seconds

Mobility model is “Random Trip Model” (avoids problems of RWP)

10-5-1-1 is a mobility scenario with mean node speed of 10m/s, speed variation of 5m/s, mean pause time of 1 second and pause time variation of 1 second

Simulation Results for Default OLSR

Adaptive OLSR vs. Default OLSR

Impact of Threshold Values

Some Analysis of Results

Adaptive OLSR also better in Packet Latency

Individual Node View

Conclusions/Future Work

Improved OLSR performance, in particular in more mobile scenarios

Idea of an adaptive protocol (nodes individually adjust their routing protocol behavior) has promise

Future Work (in increasing order of difficulty, no student right now to do any of it…. J)

Need more experimental validation

Different mobility models, in particular group mobility models

Different levels/type of offered load

Consider further changes to OLSR

TC Interval, MPR Selection, etc.

Apply these insights/ideas to other routing protocols

Consider additional avenues for adaptive behavior

A node that is currently routing multiple flows may find it beneficial to spend more efforts in routing that a node that is currently not used


	Thomas Kunz
	Systems and Computer Engineering
	Carleton University


	Adaptive Routing:
		Mobile node adjusts its routing behavior based on its environment to improve routing performance
		Adjustment is on a per-node basis, not global

	Levels/Timescales of Possible Adjustments:
		Short term: change its routing parameters based on environment
		Mid-term: select stable route with enough bandwidth during Route Discovery
		Long-term: select adequate routing protocol based on application etc


“Environment” means many things
	Radio environment: could impact route stability and throughput
	Traffic pattern: which flows, between which nodes
	Mobility: topology changes, formation/destruction of links

First step: adapt to mobility at short time-scales
	Identify mobility metric that
		Captures the mobility-driven impact of routing protocol performance
		Can easily be measured by a node, no extra hardware required
		Is true across a number of mobility scenarios (different patterns/mobility models) and independent of mobility model parameters (RWP: Max/min speed, pause time, …)
	Based on the metric, modify protocol to allow individual nodes to adapt their protocol behavior


OLSR has 4 control parameters
	Hello Interval
	TC Interval
	MPR Coverage
	TC Redundancy
Idea is to set parameters “appropriately” for network
	They then apply to all nodes and for the whole time
Our modifications:
	Monitor link breaks, define two thresholds (upper, lower)
	Choose appropriate Hello Interval
		more observed link breaks => higher mobility scenario => faster Hello
	Add states to OLSR nodes that govern processing of Hello messages/selection of MPRs
		Default OLSR, Fast OLSR, Fast Response


	Implemented Adaptive OLSR in NS2 and evaluated performance through experiments

	UM-OLSR version 0.8.8 for NS2 version 2.29
	The simulation area is 1000x1000m
	80 mobile nodes
	Default IEEE 802.11 configuration: 250 m transmission range, 11 Mbps data rate
	Traffic: 25 data sources, 4 packets/s, 64 byte packet size (CBR)
	Simulation time is 900 seconds
	Mobility model is “Random Trip Model” (avoids problems of RWP)
		10-5-1-1 is a mobility scenario with mean node speed of 10m/s, speed variation of 5m/s, mean pause time of 1 second and pause time variation of 1 second


Improved OLSR performance, in particular in more mobile scenarios
Idea of an adaptive protocol (nodes individually adjust their routing protocol behavior) has promise
Future Work (in increasing order of difficulty, no student right now to do any of it…. J)
	Need more experimental validation
		Different mobility models, in particular group mobility models
		Different levels/type of offered load
	Consider further changes to OLSR
		TC Interval, MPR Selection, etc.
	Apply these insights/ideas to other routing protocols
	Consider additional avenues for adaptive behavior
		A node that is currently routing multiple flows may find it beneficial to spend more efforts in routing that a node that is currently not used