Exploring the Energy-Latency Trade-off for Broadcasts in Energy-Saving Sensor Networks

1. Exploring the Energy-Latency Trade-off for Broadcasts in Energy-Saving Sensor Networks Matthew J. Miller, Cigdem Sengul, Indranil Gupta Department of Computer Science University of Illinois Urbana-Champaign IEEE ICDCS 2005.6

2. outlines • Introduction • Energy-efficient Communication in Wireless Sensor Networks • Probability-Based Broadcast Forwarding (PBBF) • Analytical Results • Simulation Results • Conclusion • Future Work

3. Introduction • Sensor nodes are inherently resource constrained. • Offer better reliability and performance to a sensor network application • Provide enough flexibility for a designer to choose the appropriate operation point on the resource-performance spectrum.

4. Introduction • Broadcast is useful to applications for disseminating sensor data, instructions, and code updates. • The goal is to design a broadcast protocol that allows a range of operating points from which an application designer can choose. • PBBF (Probability-Based Broadcast Forwarding), which is a MAC-layer approach and can be integrated into any sleep scheduling protocol

5. Related Work • Gossip-Based Ad Hoc Routing [5], • site percolation model • Achieving a given level of reliability requires the probability of forwarding to be beyond a threshold. • The approach does not allow an energy-latency trade-off. • PBBF protocol • bond percolation model • Two knobs, p and q, can be tuned to explore the energy-latency trade-off.

6. Energy-efficient Communication in Wireless Sensor Networks • Efficient Broadcast Protocols • Sleep Scheduling Mechanisms

7. Efficient Broadcast Protocls • SPIN protocols [6,MobileCom 1999] • Incorporate negotiation in order to avoid deficiencies of the class flooding approach. • [15][16] • Virtual infrastructure • [5,Infocom 2002][13] • To forward a message with some probability (i.e., gossip)

8. Sleep Scheduling Mechanisms • reduce energy consumption in WSNs • Active-sleep cycle • IEEE 802.11 PSM, S-MAC, T-MAC • Additional low-power wake-up radio • problem • Increasing latency • redundant packets

9. Probability-Based Broadcast Forwarding (PBBF) • PBBF exploits the redundancy in broadcast communication and forwards packets using a probability-based approach • The goal is to ensure that, with high probability, a node receives at least one copy of each broadcast packet, while reducing the latency due to sleeping.

10. The two Knobs • p • It is the probability that a node rebroadcasts a packet immediately without ensuring that any of its neighbors are active • q • It is the probability that for a given node and a given time instant when it is supposed to be asleep due to its active-sleep schedule, the node instead stays awake in the expectation that it might be a receiver of an immediate broadcast

11. PBBF exploits the redundancy in broadcast communication and forwards packets using a probability-based approach The goal is to ensure that, with high probability, a node receives at least one copy of each broadcast packet, while reducing the latency due to sleeping. Probability-Based Broadcast Forwarding (PBBF)

12. Probability-Based Broadcast Forwarding (PBBF) • PBBF exploits the redundancy in broadcast communication and forwards packets using a probability-based approach • The goal is to ensure that, with high probability, a node receives at least one copy of each broadcast packet, while reducing the latency due to sleeping.

13. Probability-Based Broadcast Forwarding (PBBF) • PBBF exploits the redundancy in broadcast communication and forwards packets using a probability-based approach • The goal is to ensure that, with high probability, a node receives at least one copy of each broadcast packet, while reducing the latency due to sleeping.

14. Pseudo-code for PBBF Sleep-Decision-Handler() • /* Called at the end of active time */ • /* If stayOn is true, remain on; otherwise sleep*/ • stayOn  false • If DataToSend=ture or DataToRecv=true • then • stayOn  ture • else if Uniform-Rand(0,1) < q • then stayOn  true --------------------------------------------------------------------------------------- Receive-Broadcast(pkt) • /* Called when broadcast packet pkt is received */ • If Uniform-Rand(0,1) < p • then Send(pkt) • else Enqueue(nextPktQueue,pkt)

15. Analytical Results • Reliability • Energy • Latency • Energy-Latency Trade-off

16. Reliability • The reliability of PBBF protocol can be analyzed using percolation model. • Percolation model, [3] • Bond percolation • Site percolation

17. Site Percolation Theory

18. Site Percolation Theory

19. Bond Percolation Theory

20. Bond Percolation Theory

21. Percolation Theory [3] • G(V,E) : an infinite connected graph • Co : the set of nodes, which can be reached by a specific node no • Θbond(Pedge) : the probability of the component Co being of infinite size so that Θbond(Pedge)=0 if Pedge<Pcbond(G)

22. Reliability (PBBF) • The probability of AB is p·q+(1-p) • p·q : A broadcasting the message immediately after reception and that B being awake at the time • (1-P) : a rebroadcast when B is awake • Each edge in the network is open with this probability. • Remark 1 (p and q for high reliability): • If Pedge=1-p·(1-q) ≧Pcbond(G), the broadcast is received at infinitely many node.

23. Reliability (PBBF) - simulator Fig.4. Threshold behavior for 90% reliability Fig.5. Threshold behavior for 99% reliability

24. Reliability (PBBF) - simulator Fig.6. Pcbond for various grid sizes Fig.7. Relationship between p and q for a given reliability level in a 30*30 grid network

25. Energy Fig.8. Average energy consumption.

26. Latency • L: the expected time between A sending the broadcast and B receiving it from A ,[4][10]

27. Latency - simulator Fig.9. Average hops traveled by an update to reach a node 20 hops from the source Fig.10. Average hops traveled by an update to reach a node 60 hops from the souce

28. Latency - simulator Fig.11. Average per-hop update latency.

29. Energy-Latency Trade-off Fig.12. Energy-Latency trade-off for 99% reliability.

30. Simulation Results • Environment parameter • assume perfect synchronization in the network • Ns-2 • The values of our parameters are based on Mica2 Mote hardware • Run time:500 sec • Each data point is averaged over ten runs

31. The impact of the q parameter Fig.13. Average energy consumption

32. The impact of the q,p parameter Fig.14. 2-hop average update latency Fig.15. 5-hop average update latency

33. The impact of the q,p parameter Fig.16. Average updates received

34. The impact of △ Fig.17. Average update latency Fig.18. Average updates received

35. Conclusion • PBBF is an efficient broadcast mechanism • PBBF provides an application designer the opportunity to tune the system to an appropriate operating point along the reliability-resource-performance spectrum.

36. Future Work • Explore how PBBF can be augmented to improve performance • The p and q parameters could be adjusted dynamically by nodes • Compare its performance with other adaptive sleep protocols.

37. Thank you