Re-routing Instability in IEEE 802.11 Multi-hop Ad-hoc Networks Ping Chung Ng and Soung Chang Liew The 4th IEEE International Workshop on Wireless Local.

Re-routing Instability in IEEE 802.11 Multi-hop Ad-hoc

Networks

Ping Chung Ng and Soung Chang Liew

The 4th IEEE International Workshop on Wireless Local Network

Overview

Motivation - Re-routing instability AODV with “don’t break before you can make” strategy (AODV_DM)

Performance Evaluation Conclusions

Motivation – (1)

A string topology

Motivation – (2)

Node 4 senses the channel to be busy since node 6 is inside its carrier-sensing range.

Motivation – (3)

Node 3 senses the channel as idle since node 6 is outside its carrier-sensing range.

Motivation – (4)

At node 4, a RTS frame or a DATA frame sent from node 3 collides with any frames sent from node 6.

Motivation – (5)

Node 3 encounters a timeout event and double the contention window size for retransmission.

Motivation – (6)

Node 6 transmits successfully and does not notice the collision at node 4.

Motivation – (7)

Node 6 uses the minimum contention window size for transmitting the next frame

Motivation – (8)

Node 6 “captures” the channel.

Although node 3 defers for a longer period before retransmission, the chance of collision at node 4 cannot be reduced.

Node 3 fails to transmit after a number of retries, it declares the link as being broken.

Motivation – (9)

The routing protocol is invoked to look for a new route.

Before a new route is discovered, no packet can be transmitted.

Therefore, the throughput drops drastically.

Motivation – (10)

There is only route from node 1 to node 7.

The routing protocol will eventually “re-discover” the same route again.

Motivation – (11)

The breaking and re-discovery of the path results in the throughput oscillations.

This phenomenon is called “re-routing instability in IEEE 802.11 multi-hop ad-hoc networks”.

Motivation – (12)

Motivation – (13)

Motivation – (14)

Throughput drops severely for the duration of 1 to 3 seconds.

It is not acceptable for real-time applications like video conferencing or VoIP.

Motivation – (15)

The routing protocol should continue to use the previous route for transmissions before a new route can be found.

AODV routing protocol is chosen as implementation details have been published in IETF RFC [11].

AODV

AODV_DM

Performance Evaluation –Simulation Setup

Each node has a droptail FIFO queue which holds up to 500 packets.

TCP Reno is used.

Throughputs are obtained by averaging over one-second intervals.

Performance Evaluation –Scenarios

A single flow in a string topologyA multiple flow in a string topology

UDP end-to-end throughput in a 7-node flow

TCP end-to-end throughput in a 7-node flow

Real-break case – Setup

Real-break case – Results

End-to-end throughput versus the number of nodes

Normalized standard deviation of end-to-end throughput versus the number of nodes

Max, Mean and Min end-to-end throughput versus the number of nodes

Multiple Flows

UDP throughputs of two 1-hop flows

Conclusions – (1)

Throughput instability problem is mainly due to a “re-routing instability problem”, rather than a binary exponential backoff mechanism.

A “don’t break before you can make” modification, which is adopted to AODV, can eliminate the instability problem.

Conclusions – (2)

Average UDP and TCP end-to-end throughputs are boosted up.

UDP and TCP throughput variations are reduced.