Improving Star Join Queries Performance: A Maximal Frequent Pattern Based Approach for Automatic Selection of Indexes in Relational Data Warehouses

Abstract

Indexing is a fundamental technique used by the administrator to reduce the cost of processing complex queries defined on a data warehouse. However, selecting a suitable configuration of indexes is a difficult problem to solve. The problem is classified as NP-hard. Automatic index selection has received significant attention in the databases field. Most works have focused on providing tools and algorithms to help data bases administrators in the choice of a configuration of indexes. Some of these works have been adapted for the data warehouse context. The idea, recently introduced, of using data mining techniques to resolve this problem remains a promising approach. In this paper, we propose a maximal frequent pattern based approach to generate a configuration of indexes from a given workload. The proposed approach was tested on APB-1 benchmark under Oracle. The results obtained show that the proposed approach generates indexes that improve the performance of the workload.