A set of application architectures and analysis processes are clustered around the technique of dimensional design. Most of these are outside the scope of this book, but we will discuss them briefly in order to put dimensional analysis in context.
Unfortunately, the field is new and terminology hasn't yet settledat one time or another most of the terms we'll discuss in this section have been used to describe the entire field. As always, I've picked a set of terms to use in the context of this book based on what seems sensible (to me) and in common use, but please be aware that other sources may not mean exactly the same thingbe sure to check.
For example, Microsoft uses the term "business intelligence" to describe the field in general, and "data warehousing" to describewell, actually they use it to describe several things. Other authors use "data warehousing" as a generic term and "business intelligence" not at all. In the context of this book, I'll use business intelligence to refer to the field as a whole.
I'll use data warehouse to refer to the physical data structures, and OLAP to refer to the applications. I'll call the individual data cubes data marts. Using this definition, a data warehouse can be thought of as the set of individual data marts used by an organization. Be aware that other authors use these terms differently. Data warehouse is sometimes used to describe the source data for the data cubes.
In the literature, the term data mining is generally restricted to the application of specific statistical techniques to the transactional data in order to extract predictive rules. Data mining is a precise mathematical field. (SQL Server 2000 provides support for some specific techniques.) It does not by any means include all of the analytical tasks for which data warehouses are used. I'll use business analysis as the generic term. (I'm sorry if you were hoping for a more exotic term, but I've never seen any benefit in neologisms when there's already a perfectly good term available.)
SQL Server Yukon adds some significant new data mining functionality to the product and a new interface for analysis services. You can expect the information in this chapter to still apply, but it will be extended in the new version.
In order to obtain a complete picture of the organization, it is usually necessary to extract data from multiple OLTP systems. Say, for example, that an organization uses two primary OLTP systems, one for manufacturing and the other for order processing. The most practical solution is to build two separate data cubes, one from each system.