Research in heterogeneous databases [Sheth & Larson 90] have provided methods to integrate disparate databases into a single unifying architecture -the federated database model. But they are limited in as much as: 1) The federated schema is non-materialized, which means that queries will have to be evaluated in the individual databases, resulting in slower response time, and 2) Data from external sources are not integrated within the federated schema. We propose to extend the federated architecture to include a data warehouse [Inmon 94, Kimball 96] modeled as a materialized view [Hanson 87] of the underlying federated schema. In addition, we employ view maintenance techniques to maintain the data warehouse against changes in the underlying operational sources. We adopt a deferredview maintenance [Colby et al 96] approach, rather than immediateapproach adopted by Stanford WHIPS project [Hammer et al 95]. This approach is preferred, because a great deal of decision-making may not require current data, but for those that require them, the model provides a mechanism to obtain them without adding too much overhead. For example, a data warehouse at a central office of a large chain of stores would like to have access to current inventory levels at individual stores, before deciding on a promotion. Similarly, access to both historical and current data of stock prices help an investment company to re-create point-in-time snapshots to help predict movements in stock prices. This approach provides the following advantages: •A unified architecture that ties the data warehouse to multiple heterogeneous databases. •Provides a method of maintaining the data warehouse as an integrated materialized view of the underlying data sources. •Provides flexible access to current data residing in the data sources. •Ease of maintenance against any change to the schema in the data source or warehouse.
V.J, Anand, "A Data warehouse within a Federated database architecture" (1997). AMCIS 1997 Proceedings. 7.