1. 1 Hierarchical Data Dissemination Scheme for Large Scale Sensor Networks Anand Visvanathan and Jitender Deogun Department of Computer Science and Engg, University of Nebraska at Lincoln Jong-Hoon Youn Department of Computer Science, University of Nebraska at Omaha ICC 2005

2. 2 Outline Introduction Hierarchical data dissemination schemes (HDDS) Basic scheme FDDD FSDD LSDD Handle node failures Simulation Conclusion

3. 3 Introduction --- background A main communication pattern is data from sensor nodes to a sink node Sink nodes may be mobile Monitor enemy in a battlefield Sensors detect enemy tanks A soldier carries a wireless communication device to collect information

4. 4 Introduction --- previous work Title: TTDD: A Two-tier Data Dissemination Model for Large-scale Wireless Sensor Networks Authors: Haiyun Luo, Fan Ye, Jerry Cheng, Songwu Lu, Lixia Zhang From: IEEE Mobicom 2002

5. 5 Introduction --- previous work mobile sink source dissemination node Record a upstream dissemination node towards the source

6. 6 Introduction --- previous work mobile sink source dissemination node Primary Agent Immediate dissemination node

7. 7 Introduction --- previous work mobile sink source dissemination node Primary Agent Immediate dissemination node

8. 8 Introduction --- previous work Advantage Source can effectively deliver data to mobile sinks Disadvantages The grid construction and maintenance per each source leads to a considerable overhead Some dissemination nodes have high overhead Some dissemination nodes does not use

9. 9 Introduction --- motivations and goals Motivations Propose efficient grid construction Balance dissemination nodes overhead Goals Scalable and efficient data dissemination Decrease power consumption

10. 10 HDDS --- assumptions Sensors are evenly and densely scattered over the sensor field All sensors know the dimension of sensor field and their location Sensors are stationary but some of sinks may be mobile Sensors use simple greedy forwarding to deliver packets to destination

11. 11 HDDS --- overview Basic scheme The non-uniform grid construction is based on sink density Three schemes for data dissemination FDDD FSDD LSDD Handle node failures

12. 12 HDDS --- basic scheme Sensor field source

13. 13 HDDS --- basic scheme Sensor field source (0,0) (l, w) (l/2-ε, w/2-ε) (l/2+ε, w/2+ε) Computed Field (CF)

14. 14 (l/2, w/2) HDDS --- basic scheme Sensor field source (0,0) (l, w) (l/2-ε, w/2-ε) (l/2+ε, w/2+ε) Computed Field (CF) The Source send a request to a randomly selected location within the CF Primary dissemination node

15. 15 (l/2, w/2) HDDS --- basic scheme Sensor field source (0,0) (l, w) (l/2-ε, w/2-ε) (l/2+ε, w/2+ε) The primary dissemination node can serve a limited number of sinks The primary dissemination node broadcasts message containing its location and coordinates of coverage area ( sensor field)

16. 16 (3l/4, 3w/4)(l/4, 3w/4) HDDS --- basic scheme Sensor field source (l/4, w/4) (3l/4, w/4) Level 1 dissemination node Primary dissemination node (l/2, w/2)

17. 17 HDDS --- basic scheme Level 1 dissemination node broadcasts message containing its location and coordinates of coverage area (1/4 sensor filed) The mobile sink receives the message, it replies with a subscription message that contains the location of its current forwarding agent Forwarding agent

18. 18 HDDS --- basic scheme Sensor field source Level 2 dissemination node Primary dissemination node Level 1 dissemination node

19. 19 HDDS --- basic scheme Sensor field source Primary dissemination node Level 1 dissemination node Level 2 dissemination node

20. 20 HDDS --- FDDD FDDD (Fully Distinct Data Dissemination) Use distinct data dissemination hierarchy for each source Each source selects a new primary dissemination node within the CF Each source may have a different set of nodes that facilitate forwarding of data

21. 21 HDDS --- FSDD FSDD (Fully Sharing Data Dissemination) Use fully sharing data dissemination hierarchy for each source The first primary dissemination node aggregates data

22. 22 HDDS --- FSDD Sensor field source Computed Field (CF) The first primary dissemination node broadcasts a message within the CF All sensors within the CF know the primary dissemination node existence sensor

23. 23 HDDS --- LSDD LSDD (Limited Sharing Data Dissemination) Use limited sharing data dissemination hierarchy for each source A primary dissemination node can serve a limit number of sources

24. 24 HDDS --- FSDD Sensor field source Computed Field (CF) sensor When this limit is reached, a new primary dissemination node is selected The new primary dissemination node then forms distinct data dissemination hierarchy

25. 25 HDDS --- handle node failures Dissemination node fails Sink node fails Forwarding agent fails

26. 26 HDDS --- dissemination node fails When a sensor node is a dissemination node, it selects one or more of its neighbors to be its backup nodes A backup nodes takes over as the dissemination node and notifies all the nodes that need to know about the change Level 1 dissemination node Primary dissemination node

27. 27 HDDS --- sink node fails If the forwarding agent does not receive sink update message within the fixed amount of time The forwarding agent transmits information to the dissemination node to stop sending data to the sink Forwarding agent Level 1 dissemination node Primary dissemination node

28. 28 HDDS --- Forwarding agent fails If the sink does not receive data from the forwarding agent within the fixed amount of time The sink starts the process of selecting a new forwarding agent Forwarding agent Level 1 dissemination node Primary dissemination node

29. 29 Simulation Ns-2 simulator 200 sensor nodes random deployment Sensor field : 2000m by 2000m 0.5% to 4% nodes are sources Each source node : 1 packet/s The number of sinks varies from 5 to 20

30. 30 Simulation The CF region is set to 20m by 20m, is located at the center of the sensor field A dissemination node can serve 3 sinks

31. 31 Simulation Energy Consumption variation by changing number of sources and sinks

32. 32 Simulation Overhead variation by changing number of sources and sinks Overhead = the number of control packet/ the number of data packet

33. 33 Simulation Delay variation by changing number of sources and sinks

34. 34 Simulation Throughput variation by changing number of sources and sinks

35. 35 Conclusion Propose energy efficient schemes for data dissemination in sensor networks The non-uniform grid can be constructed to disseminate data based on sink density