Highly-Resilient, Energy-E ffi cient Multipath Routing in Wireless Sensor Networks

1. Highly-Resilient, Energy-Efficient Multipath Routing inWireless Sensor Networks Computer Science Department, UCLA International Computer Science Institute, Berkeley ACIRI, Berkeley

2. About • Periodic low-rate floodingof data in order to allow recovery from failure. • Multipaths for energy efficient recovery • Disjoint multipath scheme • Braided multipath scheme • Braided multipaths are viable alternative for energy-efficient recovery from isolated and patterned failures

3. Introduction • Directed Diffusion • Earlier work has explored the design of mechanisms forsingle-path routing in sensor networks • To route around failed nodes, this work assumed periodic, low-rate, floodingof events that enabled local re-routing around failed nodes

4. Introduction • Multipath routing • Disjoint multipath • Braided multipath • Resilience • Maintenance Overhead • Evaluating the two mechanisms: isolated node failures and patterned failures

5. Direct Diffusion • Directed Diffusion

6. Direct Diffusion • Using directed diffusion to perform energy-efficient and robust disseminationof surveillance data samples from sources to sinks • Low rate samples • Path reinforcement • Recovery from failure along reinforcement path • The problem is low-rate flooding scheme

7. Direct Diffusion • Direct Diffusion for energy-efficient data samples from source to sinks.

8. Multipath Routing • Classic Multipath Routing usage • Using multipath routing in this paper • Primary path • Construct and maintain a small number of alternate paths (without periodic flooding) • When primary path is set up, alternate paths also sets up multipaths which data is send low-rate • No network wide flooding needed

9. Disjoint Multipaths • Small number of alternate paths that are node-disjoint withthe primary path, and with each other • How do we realize node disjoint multipaths using localizedinformation alone, and not relying on global topology information? • Primary and alternate path reinforcement • Localized disjoint multipaths are differ from idealized multipaths

10. Construction of Localized Disjoint Paths

11. Braided Multipaths • Disjoint multipaths can be energy inefficient • Alternate paths in a braid are partially disjoint from the primary path, not completely node-disjoint • For each node on the primary path,find the best path from source to sink that does not containthat node • All paths are called idealized braided

12. Braided Multipaths • Localized technique for constructing braids. • Nodes send reinforcement to route neighbours. • The alternate paths can rejoin the primary path

13. Braided Multipaths

14. Braided Multipaths

15. Qualitative Comparison • Energy/resilience tradeoffs of the two multipath schemes • The energy cost of alternate disjoint paths depends on the network density • The resilience of these multipaths to failure (isolated failures and patterned failures) • Disjoint paths give us independence, but the failure of a single node on eachalternate path results in the failure of the multipath.

16. Qualitative Comparison • By contrast, in braided multipaths, the various alternate pathsare not independent, and a combination of failures on theprimary path could sever all alternate paths • How much additional energy must one expend in orderto increase resilience by a fixed amount? • How does the energy/resilience tradeoff vary withdensity or with the extent and frequency of patterned failures? • How closely do the localized schemes approximate their idealized counterparts?

17. Evaluation Methodology • Maintenance overhead • Resilience • Failure models for which we evaluatedthe resilience of our multipath mechanisms

18. Failures • Isolated Failures • Patterned Failures

19. Details of Methodology • The idealized and localized constructions of disjoint and braided multipath in ns-2 • Uniformly distributing a number of sensor nodes on afinite plane of dimension 400 meters square • Node transmission radius: 40 meters • Density • The spatial separation between source and sink (represented by the length of theshortest-hop path between the two

20. Details of Methodology • The failure probability for isolated failures pi • the arrival rate of patterned failures is lamda p • Radius of patterned failures R • Each run of our experiment corresponded to one choice of number of nodes N and and spatial separation between source and sink d • In each run, we randomly selected alarge number of source-sink pairs separated byd hops

21. Simulation Results • Impact of failure probability on resilience: 400 nodes, 6-hop source-sink separation

22. Simulation Results • Resilience to Isolated Failures

23. Simulation Results • The impact of density and source-sink separation on resilience to isolated failure

24. Simulation Results • Resilience to Patterned Failures

25. Simulation Results • The impact of density and source-sink separation on resilience to patterned failure

26. Simulation Results • Maintenance Overhead - Density

27. Simulation Results • Maintenance Overhead – Path Length

28. Conclusions • Multipath routing for energy-efficient recovery • No need network-wide flooding for path discory on failure • Disjoint and braided multipaths are similar, but braided multipaths have about %50 higher resilience to isolated failures • It is harder to design localized energy-efficientmechanisms for constructing disjoint alternatepaths, because the localized algorithms lack theinformation to find low latency disjoint paths • Increasing the number of disjoint paths does increase the resilince but this needs higher energy cost.

29. Questions ?