Improve distributed database performance with dynamic data replication. This article introduces an adaptive algorithm that optimizes data replication in distributed systems. The algorithm dynamically adjusts the replication scheme – the set of processors storing a copy of the object – based on changes in read-write patterns, ensuring efficient data access. The algorithm aims to continuously move the replication scheme towards an optimal configuration. Demonstrated theoretically and experimentally, the algorithm can integrate with concurrency control and recovery mechanisms in a distributed database management system. The way we provide a lower bound on the performance of any dynamic replication algorithm. The results of this study are relevant for database administrators, engineers, and researchers who use distributed database systems, and more generally to scientists working on distributed computing.
While the abstract is about database systems, ACM Transactions on Database Systems also focuses on broader aspects of computer science, such as information theory and software engineering. Therefore, contextualizing the paper directly within a narrow journal scope may be limited. Considering the provided information, the paper makes a valuable contribution to database technology within the broad field of computer science.