Sufficient Statistics for Re-Optimizing Repetitive Queries

摘要

Optimizing executions of queries is the ultimate goal of arnquery optimizer. Unfortunately, due to the complexities ofrnqueries, accuracy of statistics, validities of assumptions, etc.,rnquery optimizers often cannot find the best execution plans inrntheir search spaces for the queries. In this paper, we considerrngathering statistics for re-optimization of a large and useful setrnof queries, called repetitive queries. They are likely to bernposted repeatedly in the future and deserve more optimizationrnefforts than ordinary ad hoc queries. In this research, wernidentify a minimal set of statistics, called the sufficientrnstatistics, which is sufficient and necessary to compute thernexact frequency distributions of intermediate results of allrnplans of a query. The finding of the sufficient statistics makesrnit entirely possible for an optimizer to find a plan that is trulyrnthe best in its search space for the query.