Analytics Software

Hi, folks!

It is actually a continuation of [another thread|http://forums.oracle.com/forums/thread.jspa?threadID=2265170], but I'll describe the setup anyway.

There's a fact table and six related dimensions.
Also there's a measure with the following aggregation rule defined:
Based on dimensions - checked
D_TIME (which is obviously the time dimension) - LAST
Other - SUM

Then when report is called, which includes only two dimensions (time and another typical level-based dimension) and this measure, the generated physical SQL performs joining every dimension to the fact table in an initial WITH-subquerty, even though they are not included in the final result set.
The performance is disappointing :)

I suppose, that this is caused by the Other aggregation. So the question is: is there any way to tweak logical or physical model in order to eliminate excessive joins? Or we have to put up with it and focus on RDBMS-side query optimization?

Thanks in advance!