Reading large volumes of Java objects from database

The paper explores issues that arise from storing Java objects in databases. The Java program with its classes and objects tries to maintain an object-oriented paradigm as much as possible. However, the database schema and data are not necessarily of the same object oriented mold. We report on an architecture that allows the Java program to pretend that all objects are in main memory, relieving the Java program from most database housekeeping chores. The architecture is supported by APIs to an actual database: the API can map to an object-oriented database, a relational database via JDBC, or to files using object serialization. We report on the design of these three approaches and experiences with their implementation