Modeling large scale OLAP scenarios

Abstract

In the recent past, different multidimensional data models were introduced to model OLAP ('Online Analytical Processing') scenarios. Design problems arise, when the modeled OLAP scenarios become very large and the dimensionality increases, which greatly decreases the support for an efficient ad-hoc data analysis process. Therefore, we extend the classical multidimensional model by grouping functionally dependent attributes within single dimensions, yielding in real orthogonal dimensions, which are easy to create and to maintain on schema design level. During the multidimensional data analysis phase, this technique yields in nested data cubes reflecting an intuitive two-step navigation process: classification-oriented 'drill-down'/ 'roll-up' and description-oriented 'split'/ 'merge' operators on data cubes. Thus, the proposed NESTED MULTIDIMENSIONAL DATA MODEL provides great modeling flexibility during the schema design phase and application-oriented restrictiveness during the data analysis phase.