Database replication based on group communication

Abstract

Database replication is traditionally seen as a mechanism to increase the availability and performance of distributed databases However very few of the numerous protocols that have been proposed in the literature are used in current systems There is a strong belief among database developers that existing solutions introduce too much overhead to be feasible in practice Instead current products allow inconsistencies and often resort to centralized approaches which eliminates some of the advantages of replication To resolve this issue this paper proposes a suite of replication protocols based on group communication primitives that tries to address the concerns of database designers These protocols take advantage of the rich semantics of group communication primitives and the relaxed isolation guarantees provided by most databases to eliminate the possibility of deadlocks reduce the message overhead and increase performance A detailed simulation study shows the feasibility of the approach and the exibility with which different types of bottlenecks can be circumvented.