Study of parallelism in multi-join queries

Loading...
Thumbnail Image
Issue Date
2007-04-27
Authors
Jagannathaswamy, Vivek
Advisor
Citation

Jagannathaswamy, Vivek.(2007).Study of parallelism in multi-join queries.In Proceedings : 3rd Annual Symposium : Graduate Research and Scholarly Projects. Wichita, KS : Wichita State University, p.171-172

Abstract

With relations growing larger and queries becoming more complex, parallel query processing is an increasingly attractive option for improving the performance of database systems. In this study, we exploit inter-operator parallelism to optimize the execution of multi-join queries. We focus towards the issues of scheduling the execution sequence of joins in a multi-join query and determine the optimized allocation of the processors for the execution of the joins. For scheduling the join sequences, we propose several heuristic approaches to generate the query execution plan for both the general and sequential tree join sequences. It can be inferred from our result that quality of the general join sequences is fairly close to the optimal one. Once the query execution plans are generated, we have to allocate the optimized number of the processors to the joins. In order to ensure the efficient performance of the processor allocation, we propose the concept of synchronous execution time for the inter-operator parallelism. The relationship between the first and second issue are addressed to optimize the query execution. We propose a two step approach to the query optimization for multi-join queries. We apply the join sequence heuristics to the build a bushy tree as though we build for single processors and then we allocate processors to bushy tree in top down approach through synchronous execution time for minimizing execution time.

Table of Content
Description
Paper presented to the 3rd Annual Symposium on Graduate Research and Scholarly Projects (GRASP) held at the Hughes Metropolitan Complex, Wichita State University, April 27, 2007.
Research completed at the Department of Computer Science, College of Liberal Arts and Science
publication.page.dc.relation.uri