Recent advances in Information Technology have made available voluminous amounts of data, which in turn fueled the area of Data Warehousing. In the broadest sense, a Data Warehouse is a single, integrated informational store that contains stable, point-in-time data for decision support applications. In order to facilitate faster analysis, operational data within the Warehouse is summarized (aggregated) using materialized views.

In this talk we introduce a new framework for the efficient storage and management of these views. We first present the Cubetree, a compact multidimensional index that achieves high degree of clustering, which translates to advanced query performance and several orders of magnitude faster update speeds compared to traditional relational methods. In the second part of the talk, we introduce DynaMat; a dynamic system, which automates the selection and maintenance process of the views. DynaMat constantly monitors incoming queries and materializes the best aggregates (views) subject to the available disk space. During updates, DynaMat reconciles the selection and updates the most beneficial subset of it within a given maintenance window. Our experiments demonstrate that DynaMat outperforms the optimal static selection of views and thus any practical sub-optimal algorithm that has appeared in the literature.

Back to the Spring 2000 dbchat index.