DefinePK

Abstract

In this study a new approach is proposed for partitioning and storing data cube efficiently. The new method partition the storage of the cube into multiple tables and uses a pointer table to link attribute combinations to attribute values. Experiments on synthetic and real data sets show that the required storage for the new method is about 60% of that required for flat table cube. Also the proposed approach has the advantage of fast access to the attribute values. Also it supports fast roll up and drill down without the need of any indexes for the attribute values.