Joining distributed complex objects: Definition and performance

Abstract

The performance of a non-standard distributed database system is strongly influenced by complex objects. The effective exploitation of parallelism in querying them and a suitable structure to store them are required in order to obtain acceptable response times in these database environments where performance tends to be critical. In this paper we use an hierarchical complex object model with object references and define several join operations for these complex objects which we will call tuple-objects. These join queries for tupleobjects can be categorized into materialized, functional and value-based joins. An analytical performance evaluation gives insight into which distributed storage structures for tuple-objects are most effective with regard to executing the different kinds of tuple-object joins.