Preventive Replication in a Database Cluster

摘要

In a database cluster, preventive replication can provide strong consistency without the limitations of synchronous replication. In this paper, we present a full solution for preventive replication that supports multi-master and partial configurations, where databases are partially replicated at different nodes. To increase transaction throughput, we propose an optimization that eliminates delay at the expense of a few transaction aborts and we introduce concurrent replica refreshment. We describe large-scale experimentation of our algorithm based on our RepDB* prototype (http://www.sciences.univ-nantes.fr./lina/ATLAS/RepDB) over a cluster of 64 nodes running the PostgreSQL DBMS. Our experimental results using the TPC-C Benchmark show that the proposed approach yields excellent scale-up and speed-up.