Containerized and Host-Based RDBMSs Performance Comparison

Abstract

The rapid adoption of containerization has transformed application deployment, offering significant advantages in portability and resource isolation. However, the performance implications for stateful, I/O-intensive applications like relational database management systems (RDBMSs) remain a critical concern. This paper presents a comprehensive performance evaluation comparing containerized and non-containerized deployments of two leading open-source RDBMSs, MySQL and PostgreSQL. We designed and implemented an experimental framework comprising three distinct architectural designs across nine scenarios, utilizing Sysbench to systematically measure transactions per second (TPS), queries per second (QPS), and latency. Our results indicate that, overall, container-based RDBMS deployments showed a performance advantage. Specifically, PostgreSQL consistently exhibited superior performance in a containerized environment, while MySQL's performance varied depending on the scenario. A key finding is the notable performance overhead associated with persistent storage in containers compared to nonpersistent storage. The study concludes that while containers can provide a performance benefit for RDBMSs, the realworld deployment choice requires a careful balance between performance gains and the absolute necessity of data integrity provided by persistent storage.