Knowledge-based semantic query processing—an optimal transformation approach
T.-C. Lin · University Microfilms International eBooks · 1990
Database plays a major role in all information systems, such as Transaction Processing System, Management Information System and Decision Support System. The efficiency of database (especially the relational database) depends on the Query Optimization Subsystem (QOS). The major function of the QOS is to select a cheapest access plan for a given query. There are two inherent limitations in the conventional QOS, the first is that the information used to evaluate the query comes only from the physical level. The second is that there exists only one query representation form. This dissertation develops an expert system named QUOTA (QUery Optimization Transformation Approach), which can be embedded in the relational query optimization subsystem, to relax the two limitations of the conventional technique. The major idea of QUOTA is to use Semantic Integrity Constraint rules (SIC rules) to transform the original query into several semantically equivalent (but syntactically different) queries (SEQ). It is hoped that less expensive access plans, which have been missed by conventional techniques, might be found among these. The major steps of QUOTA are as follows: (1) Select the related SIC rules from the knowledge base: Two groups of knowledge have been established in this study to reduce the search space. (2) Detect and resolve the conflicts among the SIC rules: All possible conflicts and the heuristic rules to resolve them have been identified in this step. (3) Create the best SEQ: A search technique called ABCR (All Benefit and Conflict Resolution) has been created to search the solution efficiently and effectively. A CARGO-SHIPPING relational database, in which forty five cases are used, is chosen to test the efficacy of QUOTA. The results indicate: (1) The output SEQ is always much better than the original query as far as the query processing cost is concerned. (2) Compared to all possible SEQs, the output one is always the best. (3) The output SEQ is conflict-free.