Asynchronous Ad-hoc Leader Election in Complete Networks

1. Asynchronous Ad-hoc Leader Election in Complete Networks Nolan Irving

2. Outline • Presentation of problem • Survey of current work • System description • Program description • Data collected • Conclusions

3. Problem Statement • Ad-hoc network • No existing backbone to network • Nodes are resource-poor • Battery life • Processing power

4. Problem Statement (cont.) • Leader election • At any given time, there must be at most one leader • Both links and nodes are unstable • Cannot safely assume reliable channels • Network must adapt to frequent changes

5. Uses of Ad-hoc Networks • Rapid network deployment • Combat situations • Search & rescue

6. Why Elect? • The leader is used to control requests for access to limited resources • Restoration of tokens • Grant resource requests

7. Other Research • Multihop networks • Bidirectional links • Movement-based networks

8. Flawed Assumptions • Algorithms assumed knowledge of number of participating processors • Nearly all research assumed global ordering • Link representation inappropriate to wireless networking

9. Other Problems • Maintenance costs never addressed • Addition/removal of nodes ignored • Problem increased by initialization requirement

10. Problem Description • Asynchronous network • Unknown participants • No global ordering • Broadcast communication channel • CD enabled • Complete network

11. Assumptions • Communications is a shared broadcast channel – multiple simultaneous transmissions corrupt signal • Nodes can detect a collision – likewise, the sender can detect a successful transmission • Network is single-hop – all processors can be reached with a single broadcast • A successful broadcast will reach all participating nodes

12. Program Framework • Emulate asynchronous communications using priority queue • Channel class keeps track of simultaneous communications and status • Node class handles communications requests

13. Simulation Structure

14. Program Framework (cont.) • Leader election protocol • Global ordering • Adding/removing nodes

15. Results • Times were an average of 1000 runs • Total time is listed in seconds

16. Timing Results 1

17. Timing Results 2

18. Timing Results 3

19. Comparison of Results

20. Timing Results(0.05 second message duration)

21. Summary of Results • Linear relationship between message length and election time • Polynomial growth of algorithm time and message complexity with n

22. Conclusions • Advantages • System offered a simple asynchronous protocol for leader election • Protocol allows for only one leader • Maintenance costs minimal • Handles new additions/dropped nodes easily • One of very few designs able to handle an unknown number of nodes

23. Conclusions • Disadvantages • Time not strongly bounded • Delaying technique inefficient • Will not count participating processors • Unsuited to extremely large networks

24. Sources • Fundamental Control Algorithms in Ad-hoc Networks. Hatzis, et. Al. 1998. • Leader Election Algorithms for Mobile Ad-hoc Networks. Malpani, et. Al. 2000. • Randomized Initialization Protocols for Ad-hoc Networks. Nakano, Koji and Olariu, Stephan. 2000. • Randomized Leader Election Protocols for Ad-hoc Networks. Nakano, Koji and Olariu, Stephan. 2000.

25. Questions ???