Data Quality: An Architectural Solution - High-Performance Web Databases: Design, Development, and Deployment - page 436

The corporate information factory can be extended to include the founda-
tion data store as an integral architectural construct within the framework,
as shown in Exhibit 5. The FDS functions as the official source (i.e., system
of record) for an organization's foundation data. It maintains and supplies
such data to transactional applications, the ODS, and data warehouse. The
FDS also collects and conditions external data before that data can be used
by an organization. The following sections discuss how the foundation data
store relates to the other components of the corporate information factory.

Ideally, the transactional applications should
not have their own versions of the foundation data, but should access the
centrally maintained data store. In another possible configuration, the
transactional application could make changes to a local copy of the central
store, and the changes would be applied to the central store after authen-
tication.

Transactional Applications.

The implementation of the founda-
tion data store makes the application component more integrated, which
leads to a relatively simple and straightforward integration and transfor-
mation layer.

Integration and Transformation Layer.

An ODS application usually uses the current ver-
sion of the foundation data store. Therefore, ODS applications should be
able to directly access the central foundation data store. An alternative is
to replicate a subset of the central foundation data store into the ODS en-
vironment.

Operational Data Store.

The major entities of the foundation data store become
dimensions in a data warehouse. The data warehouse contains the histori-
cal snapshots of the foundation data store. The detail contained in the
warehouse should reference the appropriate snapshot of the foundation
data.

Data Warehouse.

While doing comparative and trend analyses, users should be alerted if
different summarization algorithms are used for different time periods. For
example, if organizational hierarchy is one of the foundation data entities,
the results of summarization would depend on the hierarchy that is used in
the summarization process. Users should be given an option to view the
history based on either the same hierarchy for all time periods or different
hierarchies for different time periods.

Design and implementation of the foundation data store
should be based on a solid logical data model that is an integral part of an
enterprise data model. In the case of foundation data, translation from the
logical model to the physical design is relatively straightforward, unlike the

Data Model.