The Data Model Resource Book, Volume 3: Universal Patterns for Data Modeling

Silverston c09.tex V2 - 11/21/2008 3:16am Page 523

Data Warehouse Models 523

NOTE

We have modeled a generalized TIME dimension in the models in

Figure 9-6 (and in the star schema design that we will discuss shortly) that could

have alternatively been named ‘‘TIME BY DAY.’’ In many models, we have used

multiple TIME dimensions, each with different levels to accommodate different

granularities. Thus, depending on the circumstances and modeling style, there may

be a TIME BY DAY dimension, TIME BY WEEK dimension (having only week, month

and year), and/or TIME BY MONTH dimension (having only month and year). The

Data Model Resource Book, Revised Edition, Volume 1,providesexamplesof

these variations on TIME dimensions.

Figure 9-6 can be viewed as a stepping stone to understand the nature of

the data before creating a star schema. Many dimensional modelers skip this

type of effort and create a star schema directly from the requirements and the

source system databases. If you have the experience (and many dimensional

modelers do), you may feel confident enough to skip this type of precursor

modeling. However, we find that regardless of the level of experience, a lot of

time can be saved and many potential errors may be avoided by modeling the

data first, and then using the patterns to highlight and understand potential

complexity so you can appropriately address the issues.

A star schema is an implementation ...