THE DATA WAREHOUSE - Fundamentals of Database Management Systems - page 342

factors: their size and the portion of the company that they service (which tend to go
hand in hand), and the manner in which they are created and new data is appended
(which are also related).
The Enterprise Data Warehouse (EDW)
The enterprise data warehouse is a large-scale data warehouse that incorporates
the data of an entire company or of a major division, site, or activity of a
company. Both Smith & Nephew and Hilton Hotels employ such large-scale data
warehouses. Depending on its nature, the data in the EDW is drawn from a variety
of the company's transactional databases as well as from externally acquired data,
requiring a major data integration effort. In data warehouse terminology, a full-scale
EDW is built around several different subjects. The large mass of integrated data in
the EDW is designed to support a wide variety of DSS applications and to serve as
a data resource with which company managers can explore new ways of using the
company's data to its advantage. Many EDWs restrict the degree of denormalization
because of the sheer volumes of data that large-scale denormalization would produce.
The Data Mart (DM)
A data mart is a small-scale data warehouse that is designed to support a small part
of an organization, say a department or a related group of departments. As we saw,
Hilton Hotels copies data from its data warehouse into a data mart for marketing
query purposes. A company will often have several DMs. DMs are based on a lim-
ited number of subjects (possibly one) and are constructed from a limited number
of transactional databases. They focus on the business of a department or group of
departments and thus tend to support a limited number and scope of DSS applica-
tions. Because of the DM's smaller initial size, there is more freedom to denormalize
the data. Managerially, the department manager may feel that she has more control
with a local DM and a greater ability to customize it to the department's needs.
Which to Choose: The EDW, the DM, or Both?
Should a company have an EDW, multiple DMs, or both? This is the kind of decision
that might result from careful planning, or it might simply evolve as a matter of
management style or even just happenstance. Certainly, there are companies that
have very deliberately and with careful planning decided to invest in developing an
EDW. There are also companies that have made a conscious decision to develop a
series of DMs instead of an EDW. In other situations, there was no careful planning,
at all. There have been situations in managerially decentralized companies in which
individual managers decided to develop DMs in their own departments. At times
DMs have evolved from the interests of technical people in user departments.
In companies that have both an EDW and DMs, there are the questions of
''Which came first?'' and ''Were they developed independently or derived from
each other?'' This can go either way. In regard to data warehousing, the term, ''top-
down development'' implies that the EDW was created first and then later data was
extracted from an EDW to create one or more DMs, initially and on an ongoing
basis. Assuming that the company has made the decision to invest in an EDW, this
can make a great deal of sense. For example, once the data has been scrutinized
and its quality improved (see ''data cleaning'' below) as it was entered into the
EDW, downloading portions of it to DMs retains the high quality without putting
the burden for this effort on the department developing the DM. Development in
the other direction is possible, too. A company that has deliberately or as a matter of