An Efficient Strategy for Tiling Multidimensional OLAP Data Cubes

Computing aggregates over selected categories of multidimensional discrete data (MDD) cubes is the core operation of many on-line analytical processing (OLAP) systems. In order to support efficient computations of these aggregates in a multidimensional OLAP (MOLAP) system, a careful design of the database storage architecture must be undertaken. In particular, tiling (i.e., subdivision of an MDD cube into blocks) plays a crucial role in the overall performance of the system. Nevertheless, to our knowledge, the current MOLAP systems only provide regular tiling. In this paper we present a more efficient tiling strategy for partitioning MDD cubes in the context of MOLAP systems. We argue that, by providing explicit semantic information about the categories localization along each dimension of the MOLAP data cubes, a more accurate and efficient tiling strategy – Directional Tiling – can be implemented. Finally, we report the performance results obtained by using the described approach.