OPTIMIZATION OF EXTENDED RELATIONAL DATABASE SYSTEMS

Current relational Database Management Systems (DBMS) must be extended to function well in Engineering and Artifical Intelligence applications. Various additional functionalities have been proposed and in this thesis we study the optimization of one extended environment. Specifically, we consider the optimization of a version of the QUEL query language extended with two new features: (1) the repetitive execution of commands, and (2) the execution of relation fields in which collections of QUEL commands are stored. An extended query processing algorithm based on the original INGRES decomposition algorithm is first presented and then various modifications aiming to improve its performance are suggested. Caching of query results is also considered as another means to improve the performance of the processing engine. We analyze and suggest solutions to the various problems related to the design of a query result cache (replacement policies, invalidation techniques, etc).;In the case where all commands stored in a field are retrievals from the database, sharing of accessed data is possible. We study the optimization of processing a set of queries in detail, by deriving efficient access plans which take advantage of common intermediate results. Experimental results are also given in support of the proposed algorithms. These results show that significant savings (up to 50%) can be achieved by sharing common data.;Based on the above extensions, a relation field may contain more than one QUEL commands. Accessing such a field triggers the execution of all these commands. We present a set of tactics that can be used to reduce the cost of processing multiple commands using some interquery analysis. Special cases amenable to different kind of processing are also identified and studied.