Compare and contrast the approaches taken in the development of an EDW by Inmon’s Corporate Information Factory (CIF) and Kimball’s Business Dimensional Lifecycle.

Question

What will be an ideal response?

Answer 1

Inmon’s approach is to start by creating a data model of all the enterprise’s data; once
complete, it is used to implement an EDW. The EDW is then used to feed departmental
databases (data marts), which exist to meet the particular information requirements of each
department. The EDW can also provide data to other specialized decision support applications
such as Customer Relationship Management (CRM). Inmon’s methodology uses traditional
database methods and techniques to develop the EDW. For example, entity–relationship (ER)
modeling (Chapter 12) is used to describe the EDW database, which holds tables that are in
third normal form (Chapter 14). Inmon believes that a fully normalized EDW is required to
provide the necessary flexibility to support the various overlapping and distinct information
requirements of all parts of the enterprise.
Kimball’s approach uses new methods and techniques in the development of an EDW.
Kimball starts by identifying the information requirements (referred to as analytical themes)
and associated business processes of the enterprise. This activity results in the creation of a
critical document called a Data Warehouse Bus Matrix. The matrix lists all of the key business
processes of an enterprise together with an indication of how these processes are to be
analyzed. The matrix is used to facilitate the selection and development of the first database
(data mart) to meet the information requirements of a particular group of users of the
enterprise. This first data mart is critical in setting the scene for the later integration of other
data marts as they come online. The integration of data marts ultimately leads to the
development of an EDW. Kimball uses a new technique called dimensionality modeling to
establish the data model (referred to as a dimensional model (DM) for each data mart.
Dimensionality modeling results in the creation of a dimensional model (commonly called a
star schema) for each data mart that is highly denormalized. Kimball believes that the use of
star schemas is a more intuitive way to model decision support data and furthermore can
enhance performance for complex analytical queries.