Parallel Processing Evaluation of Path Expressions Andre Victor, Flavio Tavares, and Marta Mattoso   View Paper (PDF)   Return to Query Languages and Optimization Abstract Parallel and distributed processing are alternatives to optimize queries in Database Systems. In this work different alternatives for parallel query processing were implemented and evaluated. This evaluation aims at analyzing the potential for parallel processing of these query strategies and providing heuristics to query optimizers. The experiments were made with an IBM SP/2 parallel machine. Performance evaluation used the datasets and queries specified by the OO7 benchmark. The results indicated the best query execution strategy for different path expressions analyzed. The tests also showed a significant parallel potential for the backward, also known as pointer-based join, execution strategy. Nevertheless, the forward execution strategy, also known as naive pointer chasing, has proven its effectiveness when objects from a small collection point to objects of a large collection in the path expression, outperforming the backward algorithm both in parallel and serial executions. DiSC'01 Copyright ©2002 ACM Inc.