A Parallel Processing Strategy for Evaluating Recursive Queries

Abstract

The set of resolvents generated by a recursive intension in a lirst-order database is treated as a set of concurrent database queries. A strategy for egiciently ev,aluating these concurrent queries in a multi-processor environment is presented. The strategy combines three query processing techniques, namely, query decomposition, intermediate result sharing and data-flow and pipelined query execution to achieve a high degree of parallelism. An analytical study uses the response time for each resolvent and the execution time for a set of resolvents as a performance measure to examine the performance gain due to the data-flow and pipelined approach to query processing.