Page 90 - DCAP208_Management Support Systems
P. 90

Unit 6: Data Warehousing




          The bus architecture, recommended by Ralph Kimball, is apparently similar to the preceding  Notes
          architecture, with one important difference. A basic set of conformed dimensions (that is, analysis
          dimensions that preserve the same meaning throughout all the facts they belong to), derived by
          a careful analysis of the main enterprise processes, is adopted and shared as a common design
          guideline. This ensures logical integration of data marts and an enterprise-wide view of
          information.
          In the hub and spoke architecture, one of the most used in medium to large contexts, there is
          much attention to scalability and extensibility, and to achieving an enterprise-wide view of
          information. Atomic, normalized data is stored in a reconciled layer that feeds a set of data
          marts containing summarized data in multidimensional form (Figure 6.6). Users mainly access
          the data marts, but they may occasionally query the reconciled layer.

                             Figure 6.5: Independent Data  Marts Architecture








































          Source:  http://www.mhprofessional.com/downloads/products/0071610391/0071610391_chap01.pdf
          The centralized architecture, recommended by Bill Inmon, can be seen as a particular
          implementation of the hub and spoke architecture, where the reconciled layer and the data
          marts are collapsed into a single physical repository. The federated architecture is sometimes
          adopted in dynamic contexts where preexisting data warehouses/data marts are to be
          non-invasively integrated to provide a single, cross organization decision support environment
          (for instance, in the case of mergers and acquisitions). Each data warehouse/data mart is either
          virtually or physically integrated with the others, leaning on a variety of advanced techniques
          such as distributed querying, ontologies, and meta-data interoperability (Figure 6.7).








                                           LOVELY PROFESSIONAL UNIVERSITY                                   83
   85   86   87   88   89   90   91   92   93   94   95