Does Packet Replication Along Multipath Really Help ?

Swades DE Chunming QIAO EE Department CSE Department

State University of New York at Buffalo

Buffalo, NY 14260

{swadesd,qiao}@cse.buffalo.edu

Presentation Outline

Introduction Motivation for improved routing in wireless networks Cost issues with packet replication and selective forwarding Performance studies Results Summary and conclusion

Introduction

Possible features of ad hoc wireless networksNo centralized control

Nodes act as communicating entities as well as routersMultihop source-destination routesWide usage: required highly affordable cost required

Nodes may have limited memory and processing power Nodes may have limited mobility (e.g., in mobile ad hoc

networks)Nodes could be unattended: highly failure-prone nodes (e.g.,

in fiend sensor networks) Robust routing technique necessary

Battery operated: limited battery resource Energy-efficient routing required

Motivations for Improved Routing Existing multihop wireless routing techniques Packet replication (PR) along multiple routes (noted in [Kulik’99,

Ganesan’01]) simple but could be energy-intensive

Traffic splitting along multiple disjoint routes (D-MPR) [Lee’01,Tsirigos’01]

End node controlled – no routing flexibility at an intermediate stage

The preferred (primary) route is used, secondary routes are kept standby [Nasipuri’99, Ganesan’01] Additional energy for route maintenance Little traffic load balancing – may lead to quicker network partition

End-to-end ACK/NACK [Chen’99], or adjacent node NACK[Ganesan’01,Wan’02], or promiscuous listening [Johnson’96] based retransmission Involved flow-control mechanism, additional buffer space, transmit/receive

changeover delay, and receive power

Summary of Multipath Routing Approaches

Multipath routing approaches Disjoint multipath (D-MPR) Meshed multipath (M-MPR) [ WCNC’03]

Packet forwarding approaches Packet replication (PR) (or limited flooding) Selective forwarding [our proposed approach] Preferential routing (primary/secondary routes)

Throughput Performance Evaluation

Assumptions made in the analysis Equal length routes Regular mesh Additive white gaussian channel noise Equal failure probability of nodes (routers) For PR along meshed route, only one of possible multiple correct reception is

forwarded

An example disjoint multipath route

Meshed multipath routes considered for analysis

even odd, even odd, oddH H 2H H 2

H

Throughput Performance Evaluation (..contd.)

• 500 nodes randomly uniformly distributed in 500 m sq. area• Coverage range of each node 40 m• Gaussian channel with SNR at the receiver 14 dB Node failure probability: varied Packet size 50 Bytes 1000 data blocks (packets) per message Simulation results are within 95% confidence interval

Simulation parameters

PR vs. SF: Analytic Results

Throughput plot: 6-hop route

PR vs. SF: Simulation Results

Throughput plot: average route length (hops) 9.06

Let

PR vs. SF: Equivalent Resource Usage

T be the the packet throughput (obtained for D-MPR and M-MPR, with PR and SF, respectively

D be the message size in packets (blocks)be the # of error correcting blocks required (using FEC coding)

for successful message receptionC

Then, minimum # error correcting blocks required is: [by Ayanoglu’93]

T

TDC

)1(

Equivalent energy usage Number of transmit-receive operations, i.e.,

)( RxTxNENCD )(

Likewise, equivalent channel resource usage Number of transmit operations, i.e.,

TxCDCh )(

PR vs. SF: Equivalent Resource Usage (..contd.)

Simulated 6-hop multipath routes

PR vs. SF: Equivalent Resource Usage (..contd.)

PR vs. SF: Equivalent Resource Gain

Summary and Conclusion There have been numerous proposals that suggest packet replication (limited

flooding) along multipath as a potential routing approach in multihop wireless networks

Our original analytic studies supported by simulations show that: Packet replication approach has has higher per-packet throughput compared to

the selective forwarding approach along disjoint multipath as well as meshed multipath

However, overall performance of selective forwarding approach, when resource usage is taken into account, is superior in either multipath routing approaches.

This observation throws potential insight in routing protocol design for energy-constrained applications, such as sensor networks.

Our meshed multipath routing approach (WCNC’03) along with the observations in this paper will appear in Elsevier Computer Networks, special issue on Sensor Networks.

Thank you !