Databases Reference
In-Depth Information
query performance as one primary purpose of the design approach, not much
is needed other than some physical decisions such as physical partitioning and
indexing strategies.
In Real Life
There are many challenges in creating a dimensional model. Some of the
common ones can be observed in the big and small company groups we have
been following.
A Glimpse into Giant Co.
In this organization, members of the enterprise modeling team do most of the
modeling. The modelers are responsible for all databases, be they operational
or data warehouses. Most of their experience is with operational system design.
In addition, the modeling effort is being done without participation from the
business community. The business was involved in providing requirements,
but the interaction between them and IT ended there. Because initial involve-
ment is all that is expected from them for other systems design, that is all that
was expected of them for the DW project.
The challenge is that the modelers are not well versed in thinking dimension-
ally; and with no direct business interaction, many of the basic dimensional
modeling principles may be overlooked or ignored. The final data model is to
be presented to several select representatives from the business community.
This is mostly a formality and few changes are expected to result from this
session.
To overcome these challenges, the process to develop a data model was
refined to accommodate dimensional modeling best practices. This includes
steps to ensure participation of business representatives during the modeling
process and for reviews. This required the executive business sponsor to sup-
port the data warehouse project team's effort to get these changes put in place.
Now the enterprise modeling procedures include data warehouse - specific
steps. This helps not only this project, but also future projects too.
Insight from Agile, Inc.
At this company, everyone is used to jumping right in to get things done. In
fact, that is what happened during the first attempt to build a data mart. The
team gathered a few reports that were needed and turned them over to IT.
There was a quick database design effort, but not really a full data modeling
effort, and the design was not documented. Construction began, but as each
report was studied in more detail, this required changes to the data structures
