Combining Application-Level and Database-Level Monitoring to Analyze the Performance Impact of Database Lock Contention

Abstract

Database lock contention can severely impact application performance and limit scalability. This can be of particular importance when major modifications are made to transactional software, such as large refactorings or modernization projects. In order to assess the criticality of such modifications, it is necessary to measure the current degree of database lock contention, and attribute the effects to the appropriate sections of the application. However, current monitoring tools do not provide both application-level and database-level monitoring data with sufficient detail at the same time. In this paper, we present an approach to combine application-level and database-level monitoring to measure lock contention on a per-section basis, and present first experimental results from a prototypical implementation for PostgreSQL.