Query workload aware multi-histogram based on equi-width sub-histograms for selectivity estimations of range queries

Abstract

Query optimizer uses a selectivity parameter for estimating the size of data that satisfies a query condition. Selectivity value calculations are based on some representation of attribute values distribution e.g. a histogram. In the paper we propose a query workload aware multihistogram which contains a set of equi-width sub-histograms. The multihistogram is designated for single-attribute-based range query selectivity estimating. Its structure is adapted to a 2-dimensional distribution of conditions of last recently processed range queries. The structure is obtained by clustering values of boundaries of query ranges. Sub-histograms’ resolutions are adapted to a variability of a 1-dimensional distribution of attribute values.