Boosting Locality in Multi-version Partial Data Replication

Partial data replication protocols for transactional distributed systems present a high scalability potential, but suffer from a shortcoming of the utmost importance: data access locality. In a partial data replication setting, performance can be boosted by serving transactional read operations locally and preventing the expensive overhead of inter-node communication. In this paper we address this concern by proposing a generic caching mechanism directed towards multi-version partial data replication protocols and illustrate its application in a specific protocol, namely SCORe. Experimental results corroborate the effectiveness of the proposed caching mechanism in read-dominated workloads, where it clearly improves the system's overall throughput.