Dynamic cubing for hierarchical multidimensional data space

Abstract

Many modern applications such as sensor-based monitoring or certain business applications raise the need for real-time analysis. These applications very often operate in fast-evolving, dynamic environments. Therefore, traditional data warehouses, with their scheduled offline batch update strategy and mandatory ordered dimensions, are no longer suitable. In this paper, we present a multidimensional model for dynamic data warehousing in a hierarchical non-ordered multidimensional data space. We propose a dynamic partial cube materialisation and a tree storage structure that groups the multidimensional data in data partitions called minimum bounding spaces. Algorithms for building and maintaining the tree after a new fact is integrated and for querying the so-constructed cube are provided. We use Star Schema Benchmark and various synthetic and customisable data sets to compare the performance of our solution with the existing dynamic indexing technique. Experimental study shows performance improvement in both insertion time and queries over the data space.