Dynamic behaviour of the distributed tree quorum algorithm

Abstract

Mechanisms that ensure mutual access, replication and resilience to failures in large distributed systems can be based on quorum consensus. Effectiveness and scalability of the method selected are crucial. The aim of this paper is to present a message/time cost analysis of a distributed algorithm based on the tree quorum, which needs no global information, and has message complexity independent of the load. Results obtained for the different identifications of the requests (a requester's identification, timestamp, and group priority) illustrate the flexibility of the method The algorithm supports priority, the feature important for current middleware technologies.