I have a cube with one measure which is a distinct count measure. I took
great care in custom designing the aggregations using the Partition Manager
utility. I executed a query designed to test if it would use the aggregation
I designed. From looking at the performance monitors on that query, the DSN
Requested is matching the DSN Used, so it seems to think it's using the
aggregation. However, by looking at the Rows Read monitor, it sounds like
it's not (because it read over 100,000 rows). I also remember reading that
nonadditive measures like distinct count cannot be pre-aggregated. So my
question...

Is there any point in designing any aggregations for a cube that has nothing
but a distinct count? What if it has other measures?

Thanks for your help. This is all in SQL 2000 latest SP.

you have to also adjust the rowcounts (estimatedrowcount)of partitions
properly.



"FurmanGG" wrote:

Nick-

I assume you mean setting the "Fact Table Size" property on the cube?
There's only one partition and I'd assume it would just inherit the "Fact
Table Size" property for it's "Estimated Rows" property.

Or did you mean that you should set that property to something incorrect to
achieve a certain behavior?

We still haven't addressed the question of whether precalculated
aggregations are used in a cube with only a distinct count measure (in
AS2000). Anyone have experience with this?

Thanks for any help!

"Nick" wrote: