Hi all,

we are just trying to speed up a customer solution. During the tests we
found, that a cube with more facts than one other reacted much faster.
During the tests to find the difference between both cubes I eliminated most
of the dimensions that are different between both cubes that could be the
reason for the difference:
- One dimension with 3 parent child hierarchies
- Some virtual dimensions

The (from my point of view) fundamental difference between those two cubes
is:
- The "smaller" one (~800MB MOLAP data) has only a default measure [value]
and retrieves the data from the fact table with a dedicated dimension
[keyfigures]
- The bigger one (~1100MB MOLAP data) uses the measures dimension for 10-15
values. This cube is queried through an empty cube using "LookUpCube" to
convert the flat dimension measures into a hirarchized dimension.

The queries on the bigger one are significantly faster.

The question is:
=> Does the design of the cubes and the internal storage of data have a
correlation that can lead to those differences? <=

- Is it better to use [measures] for at least one dimension if most queries
contain at least some or sometimes all members of that dimension?

- Is there a correlation between the order of dimensions and the density of
members with data that influences query performance?

- Anything that should be known designing cubes for avoiding negative
influences on query performance? ;-)

THX
Stefan

Try to read this:

http://www.microsoft.com/technet/tre...e/AnSvcsPG.asp

Francesco De Chirico