Scalable Data Aggregation for Dynamic Events in Sensor Networks

1. Scalable Data Aggregation for Dynamic Events in Sensor Networks Kai-Wei Fan, Sha Liu, Prasun Sinha Dept. of Computer Science and Engineering The Ohio State University Sensys 2006

2. Outline • Introduction • Dynamic Forwarding over tree on Directed Acyclic Graph (ToD) • One Dimensional Network • Two Dimensional Network • Simulation Results • Conclusion

3. Introduction - Data Aggregation • Motivations • Communication cost is higher than computation cost • In-network processing reduces number/size of packets • Challenges • Dynamic events • Protocol must use low energy for long network lifetime

4. Introduction - Related Works • Static Structure • Dynamic Structure • Structure-Free

5. Data Aggregation ApproachesStatic Structure • Routing on a pre-computed structure • Doesn’t need maintain overhead • Suitable for unchanging traffic pattern • Inappropriate for dynamic event • Without or Later data aggregation a b

6. Data Aggregation ApproachesDynamic Structure • Create a structure dynamically • Has optimal aggregation • High control overhead for dynamic events a b c

7. Data Aggregation ApproachesStructure-Free (DAA, Infocom ’06) • Improve aggregation without any structure • Data aware anycast to achieve spatial convergence • Randomized waiting to improve temporal convergence • No guarantee of aggregation for allpackets

8. X Sender Pkt Sink Data Aggregation ApproachesStructure-Free (DAA, Infocom ’06) • Data aware anycast • Based on anycasting at the MAC layer for determining the next-hop node for each transmission • Randomized Waiting • Each node generating a new packet to transmit, delays it by an interval chosen from 0 toτ T=1 X X CTS Sender Sender 1 RTS(AID=1) 1 Sink Sink

9. Introduction – Motivations and Goals • Motivations • Propose a scalable structure-less protocol • Structure-free: Local data aggregation • Structured: further aggregation andGuarantee early aggregation • Goals • Low overhead of structure construction and maintenance • Suitable for dynamic event scenarios

10. Dynamic Forwarding over ToD- Basic Idea sink DAA …… …… …………………… …………………… network sink

11. Dynamic Forwarding over ToD- Basic Idea • First phase: DAA • Packets are forwarded and aggregated to the selected node(F-aggregator) • Second phase: Dynamic forwarding • Further aggregation (S-aggregator) • In one dimensional networks • In two dimensional networks

12. Dynamic Forwarding over ToD- in one dimensional networks • Assume a cell • a square with a side length • is greater than the maximum diameter of events Cell F-cluster S-cluster …… one row instance of the network …… …………………… …………………… network

13. sink F-cluster-head F-clusters Dynamic Forwarding over ToD- F-Tree • All nodes in F-clusters send theirpackets to their cluster-heads, called F-aggregators • Nodes in the F-cluster can bemultiple hops away from the F-aggregator. • Each F-aggregator then createsa shortest path to the sink

14. sink S-cluster-head S-cluster Dynamic Forwarding over ToD- S-Tree • Each S-cluster also has a cluster head, S-aggregator, for aggregating packets. • Each S-aggregator create a shortest path to the sink

15. sink F-cluster-head sink F-clusters sink s3 s4 S-cluster-head f4 a b S-cluster Dynamic Forwarding over ToD- Further aggregation f4 a b s3 s4 a b

16. sink sink S1 F2 F1 F1 b c The Event occurs in two cells and a S-cluster a b The Event occurs in two cells and a F-cluster Dynamic Forwarding over ToD- Three cases • Using DAA to determine the event span one or two cells sink F1 b The Event occurs in one cell and a F-cluster

17. An event can span at most four cells Dynamic Forwarding over ToD- in two dimensional networks A1 A2 B1 B2 C1 C2 C3 C4 A3 A4 B3 B4 S1 S2 D1 D2 E1 E2 F1 F2 A B C D3 D4 E3 E4 F3 F4 S3 S4 D E F G1 G2 H1 H2 I1 I2 G H I G3 G4 H3 H4 I3 I4 F-Clusters Cells S-Clusters

18. S-cluster S-cluster head F-cluster F-cluster head Dynamic Forwarding Rules • For packets generated only in one F-cluster, their packets can be aggregated at the F-aggregator • An event triggers nodes in different F-clusters • In the same S-cluster: aggregate at the S-aggregator • In the different two S-clusters

19. Dynamic Forwarding Rules • To guarantee the aggregation, the F-aggregator of F-cluster X forwards the packet through two S-aggregators • To firstly select the S-aggregator that is closer to the sink

20. Clustering and Aggregator Selection • Assume that sensor nodes know their physical location • Nodes in an F-cluster and S-cluster have to select an aggregator and change the role periodically • Elect themselves as cluster-head with probability based on metrics such as the residual energy • Use a hash function to hash the current time to a node within that cluster • To simplify the control overhead of the cluster head, F-cluster-head also takes the role of S-cluster-head

21. Simulation Results • Simulator: ns2 • 2000m*1200m (35 X 58 grid network) • A total of 1938 nodes • TX Range: 50m • Perfect aggregation

22. Simulation Results SPT DAA ToD OPT

23. Simulation Results • The event size is 400m in diameter • Normalized number of transmission: DAA SPT ToD OPT DAA ToD OPT

24. Simulation Results SPT DAA ToD OPT

25. Simulation Results • Event Size: 200m, 400m, 600m in diameter 600m 400m 200m for difference cell sizes

26. Conclusion • Proposed a semi-structures approach • Structure-Free Aggregation • Dynamic Forwarding on ToD for Scalability • without overhead of structure computation and maintenance

27. Thank you