Derived Horizontal Class Partitioning in OODBs: Design Strategies, Analytical Model and Evaluation

摘要

Object oriented database systems (OODBSs) have been used for advanced applications that require not only additional data modeling capabilities, but also for efficient query processing. In this paper, we apply the derived horizontal class partitioning (DHCP) technique to facilitate efficient execution of a set of queries. We apply the existing algorithms for determining horizontal class partitioning of a class for a given set of queries. Our contribution is in developing strateiges for deciding the derived horizontal class partitioning of the classes, developing an analytical cost model for determining the cost of processing queries under horizontal class partitioning, and conducting experiments to evaluate the utility of DHCP in efficiently executing the queries. The experiemntal results show that DHCP improves the overall efficiency of query execution under different object sizes and cardinality of classes, fan-out of the classes and DHCP of all classes along the class-composition hierarchy gives the best result. Further, we have classified the queries to determine the effect of DHCP on different types of queries. In particular, we find that DHCP has detrimental affect on queries that access only one class. Thus, the results from this work can be used to design object-oriented databases to efficiently process a set of queries.