A Model Driven Modernization Approach for Automatically Deriving Multidimensional Models in Data Warehouses

Abstract

Data warehouses integrate several operational sources to provide a multidimensional (MD) analysis of data. Therefore, the development of a data warehouse claims for an in-depth analysis of these data sources. Several approaches have been presented to obtain multidimensional structures from data sources in order to guide this development. However, these approaches assume that a wide documentation of the data sources is available and only provide informal guidelines to support the discovery of MD elements. Therefore, this task may become highly difficult for complex and large data sources (e.g. legacy systems). To overcome these problems, we consider the development of the data warehouse as a modernization scenario that addresses the analysis of the available data sources, thus discovering MD structures to either derive a data-driven conceptual MD model or reconcile a requirement-driven conceptual MD model with data sources. Specifically, we use concepts from Architecture Driven Modernization (ADM) in order to automatically perform the following tasks: (i) obtain a logical representation of data sources (ii) mark this logical representation with MD concepts, and (iii) derive a conceptual MD model from the marked model. Finally, we have provided a case study based on a real world project in order to exemplify the application of our approach.