Self-adaptive statistics management for efficient query processing

Abstract

Consistently good performance required by mission-critical information systems has made it a pressing demand for self-tuning technologies in DBMSs. Automated Statistics management is an important step towards a self-tuning DBMS and plays a key role in improving the quality of execution plans generated by the optimizer, and hence leads to shorter query processing times. In this paper, we present SASM, a framework for Self-Adaptive Statistics Management where, using query feedback information, an appropriate set of histograms is recommended and refined, and through histogram refining and reconstruction, fixed amount of memory is dynamically distributed to histograms which are most useful to the current workload. Extensive experiments showed the effectiveness of our techniques. © Springer-Verlag Berlin Heidelberg 2005.