Selecting views with maintenance cost constraints: Issues, heuristics and performance

In order to efficiently support a large number of on-line analytical processing (OLAP) queries, a data warehouse needs to precompute or materialise some of such OLAP queries. One of the important issues is how to select such a set of materialised views in order to minimise the total processing cost,for OLAP queries. The maintenance-cost view-selection problem is to select a set of materialised views under a maintenance cost constraint (such as maintenance time), in order to minimise the total query processing cost for a given set of queries. This problem is more difficult than the view selection problem under a disk-space constraint, because a selected view may make the previously selected views less beneficial, due to the fact that the total maintenance cost for a set of views may decrease when more views are materialised while the maintenance cost always increases under disk-space constraint. The problem has recently received significant attention. Several greedy/heuristic algorithms were proposed. However the quality of the greedy/heuristic algorithms has not been well analysed. In this paper, in a multidimensional data warehouse environment, we re-examine the greedy/heuristic algorithms in various settings, and provide users with insights on the quality of these heuristic algorithms.