Large join order optimization on parallel shared-nothing database machines using genetic algorithms

Abstract

This paper proposes the use of genetic algorithms (GAs) for optimizing the sequence of large joins execution on parallel shared-nothing database architectures. In order to measure the suitability of this method we compare the GA that we have specifically developed for this problem with previously proposed GAs. Experimental results show that our GA was able to outperform its counterparts. We also compare the performance of our GA with some known heuristics that were employed for optimizing joins in parallel queries. It turned out that for smaller number of relations, heuristics were able to produce query execution plans as good as those of GAs. However when the number of relations increases, GAs outperform heuristics.