Estimating the size of relational SPΘJ operation results: An analytical approach

Abstract

A new method for evaluating the size of intermediate results of relational query computations is here proposed. The method may be applied at compilation time for query optimization purpose and considers the results of various relational algebra operations such as Select, Project and Θ-Join (SPΘJ). Given a hierarchy (query-tree) of relational operations, the method focuses on how to infer statistics of intermediate results to be introduced in probabilistic evaluations of higher level relations, and so forth. Such statistics are statically computed for base relations and dynamically computed for any intermediate result. The hypothesis of independence and nonuniform distribution of values within attributes are here considered.