Re-Engineering Compiler Transformations to Outperform Database Query Optimizers

Abstract

Traditionally, Query Optimization and Compiler Optimizations have been developed independently. Whereas Query Optimization aims at optimizing database queries to minimize the number of disk operations, Compiler Optimizations target to maximize performance of generic executable codes. While query optimizers were originally designed for systems that needed to process large volumes of data with little main memory, the size of computer main memory has increased significantly. As a result, techniques are now being considered in the database community that have been developed in the field of compiler optimization. In this paper, we demonstrate that the converse is much more lucrative: extend compiler transformations to also target query optimization. By doing so, advanced compiler optimizations are employed as the driving force in query optimization and database systems can be on par with future complex computer architectures.