Horizontal partitioning by predicate abstraction and its application to data warehouse design

Abstract

We propose a new method for horizontal partitioning of relations based on predicate abstraction by using a finite set of arbitrary predicates defined over the whole domains of relations. The method is formal and compositional: arbitrary fragments of relations can be partitioned with arbitrary number of predicates. We apply this partitioning to address the problem of finding suitable design for a relational data warehouse modeled using star schemas such that the performance of a given workload is optimized. We use a genetic algorithm to generate an appropriate solution for this optimization problem. The experimental results confirm effectiveness of our approach.