Optimistic-Preventive Replication in a Database Cluster

In a database cluster, preventive replication can provide strong consistency without the limitations of synchronous replication. However, the original proposal [11] and its extension to partial replication [3] have performance limitations: to prevent conflicts, transactions are forced to wait a delay time before executing. In this paper, we address this problem in order to scale up to large cluster configurations. Our first contribution is an Optimistic Preventive refreshment algorithm (OptPrev) that reduces delays optimistically and prevents inconsistencies for partially-replicated databases. Our second contribution is an optimization that improves transaction throughput. We describe the implementation of OptPrev in our RepDB* prototype over a cluster of 64 nodes running PostgreSQL. Our experimental results using the TPC-C Benchmark show that our algorithm has excellent scale up and speed up.