| Resumo: | The use of materialized views is a common technique to speed up on-line analytical processing. However, the huge amount of data usually stored in data warehouses, and the complexity of their schemas, implies that only a few of the total aggregated views may be materialized. The correct selection of the materialized views is a basic condition for performance, but it is a recognized NP-hard problem. Several heuristics were proposed to the design of specific algorithms to solve that problem, being the most relevant the greedy and evolutionary ones, In this paper, we study the performance of two biological inspired algorithms applied to the cube selection problem: a genetic and a discrete particle swarm - both algorithms consider query and maintenance costs and space constraints. According to the experimental results carried on, both algorithms showed a speed of execution, convergence capacity, and consistence that allow electing them to use in data warehoust systems of medium and moderated size, being the swarm solution the one with better overall performance.
|
|---|