Graceful Degradation for Top-Down Join Enumeration via similar sub-queries measure on Chip Multi-Processor

Most contemporary database systems query optimizers exploit System-R’s dynamic programming method (DP) to find the optimal query execution plan (QEP) without evaluating redundant sub-plans. However, in the relational database setting today, large queries containing many joins are becoming increasingly common. Based on this trend, it has become temping to improve the DP performance. Chip Multi-Processor (CMP) present new opportunities for improving database performance on large queries. Based on CMP, this paper realizes the partial execution plans among the identified similar sub-queries and global execution plan among the constructed connected join pairs according to the generated partial solutions by uniform parallelizing top-down dynamic programming query optimization. Our theoretical results and empirical evaluation show that our algorithm could gracefully degrade the complexity degree for top-down join enumeration with large number of tables and impressive gains in the performance in terms of both output quality and running time.