The fact table can
contain summarized data for data items duplicated elsewhere in the warehouse, and
dimension tables can contain multiple hierarchies. As noted previously, when organizations
consolidate their data marts into enterprise data warehouses, many now
deploy a variation called a hybrid schema, a mixture of third normal form and star
schema.
Ralph Kimball, author of the widely read book The Data Warehouse Toolkit (Wiley;
see Appendix B for details), is largely credited with discovering that users of data
warehouses typically pose their queries in such a manner that a star schema, illustrated
in Figure 10-2, is an appropriate model to use. A typical query might be something
such as the following:
Show me how many sales of computers (a product type) were sold by a store chain (a
channel) in Wisconsin (a geography) over the past 6 months (a time).
The schema in Figure 10-2 shows a relatively large sales transactions table (called a
fact table) surrounded by smaller tables (called dimensions or lookup tables). The
query just described is often called multidimensional, since several dimensions are
included (and time is almost always one of them).
Pages:
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505