Materialização/utilização eficientes de visões em Data Warehouses.

Abstract

Given the complexity of many queries over a Data Warehouse (DW), it is necessary to precompute and store in the DW the answer sets of some onerous operations, so called materialized views. With this approach, one expects that queries using some of these materialized views will have acceptable processing costs, since at least some of their operations have already been computed. The complexity is then transferred to the view materialization process, the subject of much research. The main objective o f several proposed algorithms is to simultaneously materialize the views which have common dimensions (that is, common aggregate criteria). The problem with this approach is that it often imposes a storage order for the views to the detriment o f the processing cost of the queries using these materialized views. In this dissertation, we present an algorithm which allows the materialization of several views simultaneously without loosing sight of query processing costs.