Performance tuning an algorithm for compressing relational tables

Abstract

We study the behaviour of an algorithm which compresses relational tables by representing common subspaces as Cartesian products. The output produced allows space to be saved while preserving the functionality of many relational operations such as select, project and join. We describe an implementation of an existing algorithm, propose a slight modification which with high probability produces the same output, and present a performance study showing that for all test instances used both adaptations are considerably faster than the current implementation in a commercial software product.