Dynamic Evolution and Improvement of Data Warehouse Design

Abstract

Frequently changing business requirements result not only in changesto the data warehouse (DW) workload but also in changes to the schemastructure. These schema changes introduce new DW versions and so theproblem of optimization of all of its versions becomes a very challengingone.In this paper we present a new approach for dynamically evolving thedesign of the implementation schema spanning across its multiple versions.A new derivation procedure for self improvement of the design andthe performance of the model is applied in several steps. We alsodefine a system for optimization of the design of the multiversionDW schema and its particular instances and include other factors relevantfor optimization of the system i.e. view selection, vertical viewfragmentation and index selection. The self improvement of the multiversionschema design is shown by performing large set of experiments.