Databases Reference
In-Depth Information
Optimizing UDM Design
The data modeling completed during UDM creation has a significant im-
pact on both query performance and processing performance, so it is not
something to be rushed through. Even before starting to build your UDM,
you must understand the business requirements of the system under as-
sembly as much as possible. You need to have clarity regarding the goal
of the system, and that, in turn, feeds directly into creating the analysis
requirements, and what potential queries the system needs to support.
That understanding will also provide insight into what attributes the user
won't be interested in analyzing.
Every dimension, and attribute in a dimension, will demand processing
time for your UDM. In addition, because adding unnecessary dimensions
and attributes will increase the cube space, it can slow the query per-
formance too. You should use the business requirements to drive your
design, just be sure to avoid unnecessary dimensions and keep your sys-
tem compact and performant.
Fine Tuning Your Dimensions
In Analysis Services 2005, dimensions can contain several hierarchies.
Often when you create your dimension using the wizard, all the columns
in the relational table(s) are added as attribute hierarchies in support of
relational and OLAP querying. You can easily end up with a dimension
that can have hundreds of attributes. In most business scenarios, the at-
tributes within a dimension are not used in many queries. Only a subset
of the attributes might be heavily used. In this section you learn various
design techniques that will help you design the right dimension suited
to your business needs and get the optimal performance acceptable for
your business.
Choose the Right Key Attribute
Each dimension needs to have a key attribute. The relational dimensional tables
will typically have a column defined as a key column, and that is automatically in-
ferred as the key attribute by the dimension wizard. In certain relational databases,
the relationships between the fact and dimension tables and the primary and for-
Search WWH ::
Custom Search