Summary

In this chapter, we took a high-level look at dimensional databases. We saw that, while the traditional normalized databases we examined in the last section are designed for handling frequent additions and updates to the data, dimensional databases are intended primarily for ad hoc querying by end users. The data, once uploaded to the dimensional database, is rarely updated. The star and snowflake schemas are the dimensional equivalent of the entity-relationship diagrams used to define normalized databases. Star schemas define relations in terms of fact and dimension tables. Fact tables consist of two types of attributes: key attributes, which are foreign to the dimension tables, and facts, the actual data being stored.

The relations defined by a star schema can also be conceived of as a cube (technically an "n-cube"). Data cubes and star schemas are largely interchangeable; they simply represent different ways of presenting the same concepts.

Dimensional design is the analytical component of a general field known as business intelligence (also called "data warehousing" by some authors). The applications built using these techniques are sometimes called OLAP ("online analytical processing") applications to distinguish them from the OLTP applications built using entity-relationship techniques.

OLAP applications often use data mining techniques for determining trends in the underlying data. Data mining, however, is only one component of the business analysis performed using these applications.

In Chapter 7, we'll begin a more detailed examination of dimensional databases with the definition of fact tables.