Improving database performance with a mixed fragmentation design

摘要

The performance of database operations can be enhanced with an efficient storage structure design using attribute partitioning and/or tuple clustering. Previous research deals mostly with attribute partitioning. We address here the combined problem of attribute partitioning and tuple clustering. We propose a novel approach for this mixed fragmentation problem by applying a genetic algorithm iteratively to attribute partitioning and tuple clustering sub-problems. We compared our results to attribute-only partitioning and random search solution, resulting in a database access cost reduction of upto 70% and 67% respectively. We analyzed the effect of varying genetic parameters on the optimal solution through experimentation.