1 / 22

Leader Election

Leader Election. Yu Meng 09-25-2013. Outline. Basic knowledge Overview of Leader Election Complete Topology Logical Ring Topology Three Topology Latest relevant knowledge Future works References. Basic knowledge. Leader election.

huslu
Download Presentation

Leader Election

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Leader Election Yu Meng 09-25-2013

  2. Outline • Basic knowledge • Overview of Leader Election • Complete Topology • Logical Ring Topology • Three Topology • Latest relevant knowledge • Future works • References

  3. Basic knowledge

  4. Leader election • Centralized Controller greatly simplifies process synchronization • A simple point failure can limit service availability • A new controller (the leader) can be chose upon failure of the existing controller • Known to all other processes in the group • The initial of the system or a existing leader failed • The detection of failure is based on a time-out

  5. Election Criteria • Extrema Finding • Based on a global priority • Preference-based leader election algorithm • Processes in the group can vote for a leader based on a personal-preference • More general than Extrema Finding • Resulting in more complex decision-making outcome

  6. Leader election vs. Mutual Exclusion • Both try to reach an agreement for identifying a unique process • Differences: • A mutual exclusion must ensure that no process is starved, while a leader election is more concerned with the fast and successful termination of the election process • Leader election need to be announced to all processes

  7. Complete topology • Each process in the group can reach any other process in one message hop • Assumptions: • All process ids are unique and known to other process • Communication network is reliable and only the process may fail • A failure is reliable detected • Each process as a global priority and the highest-priority is elected leader

  8. bully algorithm • Extrema-finding algorithm • Process with the highest-priority process as the leader • Bully Algorithm: • Process P starts a leader election if it suspects the failure of existing leader • P sends inquiry message to nodes with higher priority • If any response then, P gives up the election and waits for higher priority node to elect itself leader • If no response then P becomes a leader

  9. Bully Algorithm • Process 4 detected leader failure and request an election • Process 5 and 6 response, then 4 stop • Process 5 and 6 each hold a election

  10. Bully algorithm • Process 6 take the response and act as the leader

  11. Logical ring topology • Easy to construct • Message initiated by node will return to itself • Indicating completion of a round of operation without the need for acknowledge • Two phases: • Initiation: One process send an election message to its successors with its ID then each process add its own ID in the forwarding message • Leader election: Message come back to initiator then the initiator announce itself as leader and broadcast to others

  12. Logical ring topology Phase 1: Initiator Phase 2: Leader election

  13. Tree topologies • A tree used for representing topological structure • Each node is considered as an autonomous entity to exchange message with adjacent nodes • A minimum-weight spanning tree (MST) are employed

  14. Tree topology • Gallager, Humbelt, and Spira’s algorithm is based on searching and combining • Starting from each node and attaching level by level till it ends up with the MST • The last node that merges and yields to the final MST becomes the leader

  15. Latest relevant knowledge

  16. Relevant research • In a distribution computing system or mobile network, leader election is a very important issue. They propose a consensus-based leader election algorithm. By analyzing the mathematic analysis and algorithm simulation results, we notice that, when a new leader is elected, the proposed algorithm guarantees a consensus be reached while at the same time reducing the number of message passing. (Chi-Chun Lo et. al., 2012) • Leader election in the presence of selfish nodes for intrusion detection in mobile ad hoc networks (MANETs). To balance the resource consumption among all nodes and prolong the lifetime of an MANET, nodes with the most remaining resources should be elected as the leaders. They justify the effectiveness of the proposed schemes through extensive experiments. (Prabir Bhattacharya et. al., 2009)

  17. Relevant research • The space complexity: The necessary or sufficient number of bits on processors to execute a leader election algorithm. Only one bit memory is sufficient for a leader election algorithm which is specific to a fixed n. A lower bound Omega(log n) on the space complexity, that is, it is impossible to construct a leader election algorithm if only log n bits are available for a processor. (Masafumi Yamashita et. al., 2008)

  18. Future works

  19. Future works • Apply the algorithm to cloud computing problem • Dynamic consolidation of virtual machine with performance and energy trade-off • Virtual machine live migration with detecting failure of physical host • Possible application to detect the failure nodes under service level agreements

  20. References • Chow, Randy, et. al.,Distributed Operating Systems & Algorithms, Addison Wesley, March 18, 1997 • R. G. Gallager, P. A. Humblet, and P. M. Spira. "A Distributed Algorithm for Minimum-Weight Spanning Trees". ACM Transactions on Programming Languages and Systems 5 (1): 66–77 • Ephraim Korach, Shay Kutten, ShlomoMoran. "A Modular Technique for the Design of Efficient Distributed Leader Finding Algorithms".ACM Transactions on Programming Languages and Systems 12 (1): 84–101 • DALAL, Y. Broadcast protocols in packet switched computer networks. Tech. Rep. 128, Dep. of Electrical Engineering, Stanford Univ., Apr. 1977 • Mohammed, N. ; Otrok, H. ; Lingyu Wang ; Debbabi, M. ; Bhattacharya, P. ,Mechanism Design-Based Secure Leader Election Model for Intrusion Detection in MANET.Dependable and Secure Computing, IEEE Transactions on, 2011, 89-103 • HUMBLET, P.A. A distributed algorithm for minimum weight directed spanning trees. Rep LIDS-P-1149, Laboratory for Information and Decision Systems, Massachusetts Inst. of Technology, Cambridge, Mass., Sept. 1981

  21. References • Hsu-Chia Cahng ; Chi-Chun Lo, "A Consensus-Based Leader Election Algorithm for Wireless Ad Hoc Networks" Computer, Consumer and Control (IS3C), 2012 International Symposium, 2012 , 232- 235 • LAWLER, E. Combinatorial Optimization-Networks and Matroids. Holt, Rinehart & Winston, New York, 1976. • LIU, C.L. Introduction to Combinatorial Mathematics. McGraw Hill, New York, 1968 • PRIM, R.C. Shortest connection networks and some generalizations. Bell Syst. Tech. J. 36 (1957), 1389-1401. • YAO, A.C.C. An O(E log log V) algorithm for finding minimum spanning trees. Inf. Process. Lett. 4 (1975), 21-23 • Andot, E. ; Ono, H. ; Sadakane, K. ; Yamashita, M. “The space complexity of the leader election in anonymous networks” Parallel and Distributed Processing, 2008 , 1-8

  22. Thank you!

More Related