Databases Reference
In-Depth Information
Figure 14.3
Example dimension table with hierarchy.
Notice that this dimension table is not normalized. Data warehouse design is
driven by efficiency of query response and simplicity. Normalization is not the driving
factor as it is in the design of databases for daily transactions. Dimensional modeling is
the dominant approach utilized for designing data warehouses. The companion topic
on logical design [Teorey, Lightstone, and Nadeau 2006] includes an overview of the
dimensional modeling approach. Kimball and Ross [2002] is an excellent and detailed
resource covering the dimensional modeling approach for data warehousing.
Figure 14.4 illustrates two date dimensions implemented as views. The
Production_Date dimension and the Repair_Date dimension are similarly structured,
with the same underlying data. Following the recommendations of Kimball and Ross
[2002], in such cases we implement one underlying table, and use a view fulfilling each
role, presenting separate date dimensions.
Figure 14.4
Example date dimensions with hierarchies.
14.3
Star and Snowflake Schemas
The dimension tables are used to group data in various ways, offering the user the free-
dom to explore the data at different levels. The measures to be aggregated are kept in a
central table known as a
fact table
. The fact table is surrounded by the dimension tables.
