Database Reference
In-Depth Information
A Typical Data Warehouse Database Design
The design of the data warehouse is similar to the OLTP you just reviewed, but its focus is different. Instead
of being concerned about normalization and the lack of redundancy, the focus is on report performance and
simplicity. A data warehouse should provide your users with a simple, high-performance repository of report
data. It should be easy to understand and consist of a minimal set of tables wherever possible.
To convert an OLTP design into an OLAP (data warehouse) design, start by identifying what reporting data
is available. Using the bottom-up approach associated with Kimball's method, focus on a particular process,
such as sales, and start building from there. It is important to try to make things very consistent so that additional
processes can be added later, in what Kimball refers to as a
bus architecture
.
Measures
Measures are the aspects of your data warehouse that are reported on, such as if a client were to say, “I want a
report on how many titles we have sold.” After taking a look at the OLTP database in Figure
4-3
, you know that the
Figure 4-3.
Selecting measure candidates
