1 / 12

Presentation by Peter Tsui for COEN 317, F/03

An Efficient and Fault-Tolerant Solution for Distributed Mutual Exclusion by D. Agrawal, A.E. Abbadi. Presentation by Peter Tsui for COEN 317, F/03. Goals. Fault-Tolerant - sites, links, and network partitioning Low Communication Cost Distributed Solution. Background – Other Solutions.

marla
Télécharger la présentation

Presentation by Peter Tsui for COEN 317, F/03

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. An Efficient and Fault-Tolerant Solution for Distributed Mutual Exclusion by D. Agrawal, A.E. Abbadi Presentation by Peter Tsui for COEN 317, F/03

  2. Goals • Fault-Tolerant - sites, links, and network partitioning • Low Communication Cost • Distributed Solution

  3. Background – Other Solutions • Primary Site • Distributed (Lamport) • Majority • Maekawa • Token-Based (Raymond)

  4. 2 Ideas • Coterie, C, is a set of sets or quorums. • Intersection Property – If g and h are quorums in C, then g and h has non-empty intersection. • Minimality Property – There are no two quorums g and h in C such that g is a superset of h. • Use Logical Structure – Organize sites into a logical (complete) binary tree (easy extension to n-ary tree)

  5. Tree Quorum TQ(T) Construction • If AVAIL(T->Node) return • T->Node U TQ(T->Left-C) or • T->Node U TQ(T->Right-C) • Else return • TQ(T->Left-C) U TQ(T->Right-C) • If either of above is empty, error

  6. Example No fail : {1,2,4}, {1,2,5}, {1,3,6}, {1,3,7} {1} fail : {2,4,3,6}, {2,4,3,7}, {2,5,3,6}, {2,5,3,7} {2 or 3} fail : {1,4,5}, {1,6,7} {1, 2} fail : {4,5,3,6}, {4,5,3,7} {1,2,3} fail : {4,5,6,7} 1 2 3 4 5 6 7 2-level complete binary tree

  7. Correctness • The set of Tree Quorums satisfy intersection and minimality properties of coteries. • Proof by inducation of level of tree. • Induction Step: Consider binary tree of level k+1. Any tree quorum is from one of following 3 classes:- • {s1} U {members from quorum set of left subtree} • {s1} U {members from quorum set of right subtree} • {members from quorum set of left subtree} U {members from quorum set of right subtree}

  8. Discussion • Best case tree quorum size is (ceiling function on) log n (base 2). • Worst case tree quorum size is (ceiling function on ) (n+1)/2. • If {1,2,4} fail, no tree quorum possible. • When {3,5,6,7} fail, a majority, tree quorum still possible. • When less than log n nodes fail, tree quorum always possible.

  9. Message Cost f = fraction of tree quorum that include ROOT of tree Recurrence relation of cost of tree quorum at level (l+1) in terms of l.

  10. Availability Availability of Tree Quorum indistinguishable with Majority Quorum when p, probability of site availability, > 0.75. Availability is only inferior when p is between [0.5, 0.75].

  11. Some Applications • Mutual Exclusion – send request to a tree quorum, get all replies to enter CS, send relinquish when done. Use Inquire/Yield to break deadlocks. • Replication – use tree quorums for both read and write quorums. Get data with highest version number.

  12. Conclusions • Tree Quorum provide log n messaging cost in best case and (n+1)/2 in worst case. • Tree Quorum provides fault tolerance up to about log n down sites in worst case. • Tree Quorum availability is comparable with majority alogirthm when p > 0.75.

More Related