View Selection in OLAP Environment

Abstract

To select some “valuable” views for materialization is an essential challenge in OLAP system design. Several techniques proposed previously are not very scalable for systems with a large number of dimensional attributes in the very dynamic OLAP environment. In this paper, we propose two filtering methods. Our first method, the functional dependency filter, removes views with redundant summary information based on functional dependencies among the dimensional attributes. The second method, the size filter, is based on the view size to filter out any view that can be either derived from another small materialized view or has almost the same number of tuples as another materialized view from which it can be derived. More over, all useful views are selected by these two view filtering methods, other existing view selection methods can still be applied on the remaining views to further reduce other possible non-essential views from systems. We conduct performance tests to compare our method with other existing methods. The results show our method outperform the others.