Abstract:

Database replication is a well-known technique for increasing performance （response time, throughput, etc. ） and availability in distributed databases. Over the past few years, a database replication protocol based on group communication has been widely used in constructing practical replicated data   bases for its advantages in easier and more flexible implementation with good performance compared to traditional replication protocols. To keep the con   sistency of all replicas on sites is a key problem of the research in database replication. This paper discusses the problem of how to regain the consis  tency of replicas after the recovery of the failed sites or adding new sites. A recovery protocol for distributed replicated databases is provided. Comb ined with the replication protocol based on group communication, the protocol can make the replicated system recover correctly after a site, which is co   mpletely new or has recovered from a failure, rejoining the system, while the whole system can still process transactions without being interrupted. We   give a theoretical correctness proof of it, and also prove that it is suitable for the replicated databases which are distributed on an unreliable netwo rk

Key words: database replication, database recovery, group communication