Abstract:
Real data warehouse workloads are often too large and complex to be directly optimized using the algorithms proposed in the literature for view materialization and indexing. In this paper we propose the idea of profile as an instrument for summarizing the workload features in order to help the designer to make the right choices. The ability of the profile to characterize a workload is then exploited to move backward using it as an input for an algorithm that generates a set of queries presenting the desired features. The algorithm proposed is finally used for creating the workloads necessary for testing the correspondence between different profiles and the results of optimization.