Data Warehouse Architecture for DSS Applciations

Abstract

A data warehouse facilitates the integration of disparate operational databases in an enterprise into a single store. The warehouse then provides knowledge workers with easy access to historical, summarized and other forms of aggregated data. A major flaw in present warehouse architectures is the de-coupling of the warehouse database from its underlying operational databases. This creates two problems: the difficulty in updating the warehouse and the inability to provide users with a drill-down feature from warehouse to current data. We propose an architecture that is based on a federated database system extended to incorporate a materialized warehouse. A federated database provides the basic mechanism to tightly couple heterogeneous databases, which in this case are the operational and warehouse databases. The warehouse database is a special component in the federation, made up of historical data, external data and materialized views of the operational data. This approach enables users to access historical and current data when required, and provides a method of maintaining the warehouse as an integrated view of the underlying operational data sources.