Dynamic reorganization of object databases

We consider the problem of optimizing the number of I/O operations by means of an effective clustering, with the objective to overcome deficiencies of existing solutions. Our main objective is to dynamically adapt physical database organizations, online, according to the access patterns of the users without adding significant overhead on processing workload. We propose a clustering schedule that may be convenient for any object database management systems (ODBMS); it combines flexibility, dynamics, and efficiency. The paper describes relative partial validation of our approach using Versant.