A Cost Model for Clustered Object-Oriented Databases

Abstract

Query processing is one of the most, critical issues in Object-Oriented DBMSs. Extensible opt,imizers with efficient, search strategies require a cost model to select the most efficient execution plans. In this paper we propose and partially validate a generic cost-model for Object-Oriented DBMSs. The storage model and its access methods support clust,ered and nested collections, links, and path indexes. Queries may involve complex predicates with qualified path expressions. We propose a, method for estimating the number of block a,ccesses to clustered collections and a paramet,erized execution model for evaluating predicat,es. We estimate the costs of path expression traversals in different cases of physical clustering of the supporting collections. Thr model is validated through experiments with the 02 DBMS.