A New Multipath Routing Protocol for Ad Hoc Wireless Networks

1. A New Multipath Routing Protocol for Ad Hoc Wireless Networks Amit Gupta and Amit Vyas

2. Outline • Motivation • Intuition and Previous Work • Problem Statement • Analysis • Simulations • Conclusion June 5, 2002

3. Motivation • Ad Hoc Wireless networks • Very quickly deployable • Less costly than infrastructure networks • Applications • Internet connectivity in buildings • Connecting computers at a conference • Video-conferencing • Wireless sensor networks • Military communications and emergency services June 5, 2002

4. Motivation • Interactive Applications place stringent demands on delay requirements (e.g. telnet or multimedia communication) • Lower delays might be more important than bandwidth used • Can we achieve lower delays by sending same packet over multiple paths ? • What is the trade-off between decrease in delays and increase in bandwidth ? June 5, 2002

5. Outline • Motivation • Intuition and Previous Work • Problem Statement • Analysis • Simulations • Conclusion June 5, 2002

6. Intuition • Send same packets along multiple paths • Lowers end-to-end probability of losses • In turn reduces expected number of retransmissions and number of timeouts • Hence decreases expected end-to-end delay • More relevant for wireless networks • They are more lossy than wired networks • But increases bandwidth requirements • Let us try to quantify this trade-off June 5, 2002

7. Previous Work • Old work • Dispersity Routing – Maxemchuk (1975) • Efficient Dispersal of Information – Rabin (1989) • More recently • Apostolopoulos (2001) • Split a video into multiple streams and send different streams over different paths to achieve better quality • Liang, Steinbach, Girod (2001) • Real-time voice communication over the internet using packet path diversity. • Experimental setup in which they send packets over two paths using a relay server. • Also simulated two CBR voice streams via two paths using ns simulator June 5, 2002

8. Outline • Motivation • Intuition and Previous Work • Problem Statement • Analysis • Simulations • Conclusion June 5, 2002

9. Problem Statement • But no one has looked at • Impact of using extra paths on bandwidth • Trade-off between reduction in delay and extra bandwidth used • Optimal number of redundant paths • Depends on loss probability, time out values, number of hops • Application to routing protocols June 5, 2002

10. Problem Statement • Outline of Work • Analysis of delay-bandwidth tradeoff • Study of loss models for wireless networks • Impact of various parameters on the optimum choice • Study of current Ad Hoc Wireless Routing Protocols • Design of a new protocol • Simulations using ns-2 network simulator June 5, 2002

11. Outline • Motivation • Intuition and Previous Work • Problem Statement • Analysis • Simulations • Conclusion June 5, 2002

12. Analysis • Loss model for wireless networks • Nguyen et. al. (1996) • Trace-based approach for modeling wireless channel behavior • For AT&T WaveLAN, average packet error rate of 2-3 % • Improved two-state model (error, error-free) • Two & Three segment curves respectively for distribution • Konrad et. al (2001) • Markov based channel model • Algorithm to divide trace into stationary components • These models are very complicated and need very specific parameters • Simple probability model is good enough for our study • Burst losses and outages will only make our results better June 5, 2002

13. Analysis • Important parameters • Loss probability over a link • Timeout for end-to-end retransmission • Number of hops • Number of paths • Additional delay of alternate path(s) • Metrics • Expected end-to-end delay • Expected bandwidth used • Delay-bandwidth product June 5, 2002

14. Analysis • Variation of number of paths June 5, 2002

15. Analysis • Variation of number of paths June 5, 2002

16. Analysis • Variation of number of hops June 5, 2002

17. Analysis • Variation of number of hops June 5, 2002

18. Analysis • Variation of additional delay of alternate paths June 5, 2002

19. Outline • Motivation • Intuition and Previous Work • Problem Statement • Analysis • Simulations • Conclusion June 5, 2002

20. Current Ad Hoc Wireless Routing Protocols • Table-Driven Routing Protocols • Destination-Sequenced Distance Vector Routing (DSDV) • Clusterhead Gateway Switch Routing (CGSR) • The Wireless Routing Protocol (WRP) • Source-Initiated On-Demand • Ad Hoc On-Demand Distance Vector Routing (AODV) • Dynamic Source Routing (DSR) • Temporally Ordered Routing Algorithm (TORA) • Associativity-Based Routing • Signal Stability Routing June 5, 2002

21. Simulations • To validate the analysis results • Used the ns-2 simulator with wireless extensions • Modified DSR protocol to allow multiple paths – Dynamic Multipath Source Routing (DMSR) • Simulation Setup • Interested in study of delay-BW tradeoff for end-to-end error-recovery protocols (e.g. TCP) • Effect of number of hops and network load • Two scenarios – deterministic 16 nodes and uniform randomly distributed 30 nodes June 5, 2002

22. Simulations • Scenario of 16 nodes • Relatively low network load June 5, 2002

23. Outline • Motivation • Intuition and Previous Work • Problem Statement • Analysis • Simulations • Conclusion June 5, 2002

24. Conclusion • Use of multiple paths for achieving lower expected delays • Studied effect of various parameters on the delay-bandwidth trade-off • Determined optimal number of redundant paths • Implemented scheme as DMSR Wireless Routing Protocol • Validated results using simulations June 5, 2002

25. THE END June 5, 2002