Performance evaluation of new clustering algorithm in object-oriented database systems

Abstract

It is widely acknowledged that clustering is a cornerstone for the performance of database management systems. In particular, object-oriented databases have properties which may compromise the effectiveness of a clustered structure, i.e., object updates and multiple relationships. The clustering problem consists in finding a partition of the set of the objects in the database. Our approach is similar to the one adopted in Schkolnick [1] for the hierarchical model. When objects are grouped together, it is desirable that these objects be accessed together in the future. We have an operational approach: we will evaluate the benefits of our clustering strategy and show how to dynamically adapt it.