Collateral evolution of applications and databases

Separating the evolution of an application from the evolution of its persistent data, or from the evolution of the database system used to store the data can have collateral effects, such as data loss, program failure, or decreased performance. In this paper, we use empirical evidence to identify challenges and solutions associated with the collateral evolution of application programs and databases. We first perform an evolution study that identifies changes to database schemas in two popular open source applications. Next, we study the evolution of database file formats for three widely-used database management systems. We then investigate how application programs and database management systems cope with these changes, and point out how collateral evolution can lead to potential problems. Finally, we sketch solutions for facilitating and ensuring the safety of application and database evolution, hence minimizing collateral effects.