Analyzing the adoption of database management systems throughout the history of open source projects

Analyzing the adoption of database management systems throughout the history of open source projects

연구 분야: Databases

학회: Empirical Software Engineering

The appropriate selection of DBMSs (Database Management Systems) is relevant for the success of modern software applications. Relational DBMSs are popular for structured data management, while non-relational systems, such as NoSQL databases, have gained traction for handling unstructured data and scaling in dynamic environments. These varying DBMS characteristics have led to an increasing trend of combining multiple systems within a single application to meet diverse requirements. However, existing work does not analyze whether DBMS are replaced or used together in a broad scope. This paper presents an empirical study on DBMS usage across 362 popular open-source Java projects hosted on GitHub. Our analysis focuses on the most widely adopted DBMSs, both relational and non-relational, as ranked by the DB-Engines website. By examining DBMS integration patterns, stability, and migration trends, we aim to uncover insights into the factors driving DBMS choices in real-world applications. We investigated DBMS popularity, usage stability, migration patterns, synergy among DBMS, and the role of Object-Relational Mappers (ORMs) in DBMS interactions. We applied heuristics to detect DBMS presence, tracked usage trends over time, and analyzed the coexistence and replacement of different systems. We also examined ORM frameworks to understand their impact on DBMS management and query-building practices. Our findings reveal that MySQL and PostgreSQL are the most popular DBMSs, although some projects replace them with other DBMSs. While certain popular DBMSs (e.g., Redis, MongoDB) usually stay in the project after they are introduced (and therefore their adoption is stable), others (e.g., HyperSQL) are frequently replaced as project requirements evolve. We also observed patterns of polyglot persistence, where multiple DBMSs coexist to handle varied data types. Notably, Informix is a relational DBMS designed to handle real-time data processing and is always used with other DBMSs. Additionally, we identified ORM usage trends that facilitate database interactions and mitigate migration complexities. These insights contribute to a broader understanding of DBMS adoption, providing valuable guidance for developers and architects in selecting and managing database infrastructure over time.