I have a MOLAP cube based on historic data that never
changes. The cube's other dimensions, except for time,
change quite frequently due to realignments in things like
the company's "organization chart" and "product department
groups". All these changes are Type I, and are really just
a regrouping of the children under different parents.

If I define these dimensions as CHANGING, it appears that
any such regroupings can be accommodated by simply
reprocessing the dimensions themselves, and not the cube.
Apparently, Analysis Server is not forming any
aggregations along these dimensions.

The bottom line is that the underlying fact table in the
database is no longer required in these situations. Is
this just a bit of luck in my current environment, or is
there a predictable way to guarantee this?



Thanks

Hi Tom,
It is always a good idea to retain the fact table.
This would help in case you or your organisation need(s) to make
design changes to the cube at a future date.
The assumption as I read your post is that there would never be a need
for additional dimensions or additional measures....which wouldn't be
a good assumption to have. You might encounter needs for new
dimensions or measures etc that would need you to access the fact
table afresh.

- Priya



"Tom K" <anonymous (AT) discussions (DOT) microsoft.com> wrote

A sound principle - but the question is: is it _required_
to retain the underlying fact table in the RDB?