Improving Materialized View Selection Under Storage Constraint

Data warehousing and on-line analytical processing (OLAP) have increasingly become a center of the database industry. The ability and the insurance that they give to companies for the success of their activities are greatly improved.An important issue in data warehouse development is the selection of a set of views to materialize in order to accelerate Online analytical processing queries, given certain space and maintenance time constraints. On the other hand, the materialized views (MV) selection problem is an NP-complete; the number of possible summary-views in the multi-dimensional (MD) model increases explosively with the increasing number of dimensional attributes. The high storage cost and computation cost make it unfeasible for any system to materialize all of these possible views. Thus, precising the summary-views for materialization becomes an essential challenge in OLAP research.To solve this problem of summary-views explosion and storage cost, we supply an improved greedy algorithm for selecting best views for materialization based on their weights.Given a space constraint for the materialized set, the algorithm picks a view with the maximum Benefit Per Unit Space (BPUS) to materialize from the candidate view set. After a period of time (precisely at the maintenance time), and giving a new query frequency for each view, the algorithm determine the view to be removed from the materialized view set. Logically speaking those view removed are views that are less useful to the data warehouse. The final step is to fill the space left by the removed view with the other new one until the optimal materialized view set is reached.During the experiment of this algorithm, we observe that the after a period of time, some materialized view weights (BPUS) become very thin while other candidate view from the candidate view set are having more important. This new method provides a way to solve this problem of useless materialized view. We think this work will contribute greatly in the improving the efficiency of materialized views usage in the data warehouse usage.