Query processing remains one of the most important challenges of object oriented database systems (OODBSs). A cost model for query processing is very useful in performing the physical query optimization. Horizontal class partitioning (HCP) is a technique for reducing the number of disk accesses for executing a given set of queries by minimizing the number of irrelevant object instances accessed. Moreover, its importance in reduction of query execution time has been widely acknowledged. However, existing cost models for query processing in OODBSs do not take into consideration the HCP criteria. In this paper, we present an analytical cost model for query execution for unpartitioned and horizontally partitioned classes. This cost is necessary for deciding whether to use HCP or to keep all classes unpartitioned. The effect of HCP upon the query execution process is analyzed. Key Words : Object-oriented database, horizontal class partitioning, cost model, query processing, query optimizat...

Derived horizontal fragmentation is one of the main database distribution design techniques. Unlike primary horizontal fragmentation, the decision of derived horizontal fragmentation is not straightforward. In the literature, in the context of the relational model, derived horizontal fragmentation of a member relation is achieved by performing semijoins with fragments of one of its owner relations, which is chosen in an ad hoc manner without evaluating the system performance. Similar approaches are found in the literature for the object oriented data model. We note that fragmentation and allocation are often considered separately, disregarding that they are using the same input information to achieve the same objective, i.e. improve the overall system performance. This paper addresses derived horizontal fragmentation and allocation simultaneously in the context of complex data model. The core of the paper is a heuristic approach to derived horizontal fragmentation, which uses a cost model and is targeted at globally minimising costs.

Operations are effective: An attempt to apply machine learning to database parallelization by