Data Synchronization and Resynchronization for Heterogeneous Databases Replication in Middleware-based Architecture

Currently more and more web applications and telecommunication applications are required to be separation with database layer. A general solution is to employ a middleware-based architecture, including an application client tier, a database front end (DB-FE) tier, also called middleware tier, and a database backend (DBBE) tier. This paper mainly focuses on the issues of data synchronization and resynchronization in the case of failures for the architecture. Firstly, a synchronous update everywhere replication model and a replication 2-phase commit protocol (R2PC) are discussed, which can increase update success ratio of replication transactions through relaxing the commit constraints, and enhance DB-BEs’ availability using transaction retry and discard mechanisms. Then, a novel method using request logs is suggested for data resynchronization. The method only records the missing transaction requests for the unavailable DB-BEs in the DB-FEs and links the logs belonging to a site together, to reduce the resynchronization time. Moreover, the method doesn’t interrupt the normal transaction processing in the DB-BEs, so system throughput using request logs is higher than using transaction logs. Experiences validate that the suggested methods have better performance.