Virtual Cube in FrameWork

Hi Experts !
I am working in a new company.
When they were witn Cognos 7, they had two cubes ( sales and purshasing)
When they moved to Cognos 8.4 they decided to build virtuals Cubes whinin Framework.
The purpose was to avoid building a cube for 8 to 10 hours every day.

Now, with Analysis Studio, it takes a very long time for the users to get their data (much too long).

I have ideas how to improve performance of a cube via Transformer,
but what about Framework.

Thanks in advance !!

System :
Windows Server 2003
Service pack 2
SQL server 2008

Copied from you know who from you know where, and I quote:

"A DMR model will never out-perform a true cube, but in many cases it can still be an acceptable alternative. As each analysis action triggers an SQL on the database the first thing is to check whether this is still acceptable. If the optimizer cannot create a optimal path queries can become bottle-neck.
Having the right indexes in the database and fresh statistics are a must.
With very large datasets it may become important to define indexed views that store pre-aggregates synchronized with the base tables.

Modern databases on dedicated servers should not have much trouble with a couple of million fact records, which begs the question how large your tables really are."