Algorithm choice for multiple-query evaluation

Abstract

It is pointed out that traditional query optimization concentrates on the optimization of the execution of each individual query. More recently, it has been observed that by considering a sequence of multiple queries some additional high-level optimizations can be performed. Once these optimizations have been performed, each operation is translated into executable code. It is shown that significant improvements can be gained by careful choice of the algorithm to be used for each operation. This choice is not merely based on efficiency of algorithms for individual operations, but primarily on the efficiency of the algorithm choices for the entire multiple-query evaluation. An efficient procedure for automatically optimizing these algorithm choices is given.<>