Efficient Materialized View Selection for Multi-Dimensional Data Cube Models

1. Introduction

Decision Support Systems (Harinarayan et al 1996) are used by all business enterprises to gain competitive advantage in current market scenario. All the decision support systems are supported by huge data warehouses at the backend. Complex queries run on data warehouses and results achieved. Query response time is one of the major factors affecting the quality of data warehouses. A number of query optimization (Rafanelli 2003; Parimala and Kumar 2002) techniques exist of which one such technique involves greedy selection of views that defines the order of view selection. The greedy approach (Harinarayan et al 1996) selects the view that has maximum cost benefit among all the views not selected so far. The greedy approach aims at selecting a fixed number of views to be materialized irrespective of space constraint.

The authors aim to select materialized views following a refined greedy approach. At each pass one view is selected that gives maximum cost benefit. When more than one views with same maximum cost benefits exist, then the choice of views is made by selecting the view that gives maximum cost benefit in next subsequent pass. If multiple views with same maximum cost benefits exist in the subsequent pass, then that view is selected corresponding to which a view has maximum cost benefit and minimum occupied space in the subsequent pass. The proposed approach considers cost benefits along with space occupied in greedy selection of views, makes use of forward referencing and is capable enough to give better results than existing greedy approach. The results achieved using refined greedy approach is then used for view selection under space constraints using knapsack implementation.