|
Abstract
Many modules, such as query optimizers, in a DBMS require an accurate
estimate of the selectivity of queries. Histograms are one of the most
commonly stored statistics for selectivity estimation. They take
reasonably small amount of space and their estimates are quite accurate.
A histogram approximates a data distribution by grouping the data values
into buckets, and approximate the actual distribution by the statistics
stored in each bucket. Different types of histograms are proposed in the
literature. In this talk we will discuss a taxonomy of existing
histograms for signle-dimensional data. Few extensions of histograms for
multi-dimensional data will also be discussed.
Read the Presentation
Slides...
Referred Papers
|