SLEMAS: An Approach for Selecting Materialized Views Under Query Scheduling Constraints

Materialized views are one of the most popular optimization techniques selected during the physical phase to speed up query processing in traditional and advanced databases. Their selection has been proven to be NP-hard. As a consequence large panoply of heuristics has been proposed to find near optimal solutions. Usually, the selected materialized views are whole life disk resident and their presence is not calling into question. Note that view maintenance can cause significant amounts of CPU and I/O usage, which can be detrimental to performance in a write-intensive database application. Typically materialized views are stored on disk; however with big number of queries, there are situations where not all a good candidate views will be selected. As a consequence, their dynamic selection becomes a necessity. In this paper, we address the problem of materialized view selection by considering the query scheduling. We first review the most important existing work on static and dynamic view selection. A formalization of the problem of view selection considering the re-ordering of a large number of queries is given. A system, called SLEMAS, playing the role of a generic advisor is described. Finally, intensive experiments are conducted to compare the efficiency of our system regarding the most important state of art algorithms