Database Reference
In-Depth Information
The format of the unique name will vary depending on the type of hierarchy and
various property settings on the dimension, but as we can see from the preceding
examples, the uniqueness of a a unique name usually derives from values in the
columns you've used in the
KeyColumns
property of the attribute hierarchy. If a Type
I change alters one of these values then that will carry through to a change in the
unique names of one or more members on hierarchies in the dimension. If a saved
query references the old version of the unique name, one of two things will happen:
•
If we have set the
MDXMissingMemberMode
property of the dimension to
Ignore
(and this is what the
Default
option means, too), then any queries
which refer to the old version of the unique name will simply ignore that
member. This would result in rows or columns disappearing from queries
or reports with no warning.
•
If we have set
MDXMissingMemberMode
to
Error
, then any queries that refer
to the old version of the unique name will raise an error when they are run.
At best, this will mean you have to manually fix the MDX query so that the
old version is replaced with the new version; in some Analysis Services client
tools, it will mean you have to delete the entire query or report and recreate
it from scratch.
Type II SCDs
Type II SCDs represent a more difficult problem: unlike some other OLAP servers,
Analysis Services has no explicit support for different versions of the same dimension,
and it's up to us as developers to decide how we should design our dimension to
handle such changes.
When a Type II change takes place, a new row is added to a dimension table and this
in turn results in the addition of a new member on the key attribute of the Analysis
Services dimension. This has some positive implications: if we only allow Type II
changes on a dimension, because we are only adding new rows and not updating
existing rows, we can do a
Process Add
on our dimension which will be faster than
a
Process Update
, and we can set all our attribute relationships to
Rigid
so we
do not need to rebuild any aggregations after processing. However, modeling the
attributes and attribute relationships on a Type II SCD can require some thought.
Search WWH ::
Custom Search