Cube Solve Time when using MAX Aggregation Operator

Question

Hello,
We have created a cube to implement the count distinct measure we need.
The cube contains only one measure (COUNT) and uses the MAX operator to aggregate across all other dimensions except for the one we want to count (which uses the SUM operator). We have set the precompute percent to 60% for the bottom partition and 0% for the top partition. The cube is compressed.
The problem is that the SOLVE step for a partition when performing a COMPLETE cube build, seems to be taking a very long time and is taking up huge amounts of TEMPORARY tablespace.
We have succesfully created another cube with the same dataset which uses the SUM operator across all dimensions.
This cube build was completed in a reasonable amount of time even though we had 5 stored measures and 80% aggregation for the top partition.
Is this behaviour expected when using MAX operator?
Thank you,
Vicky

Answer 1

Thank you, David.
As you said we are using mixed operators because we are doing a distinct count.
We will try setting the precompute percent to 35%,although I'm a bit worried about the query performance in this case.
Neelesh, I think that Atomic Refresh was set to TRUE during the last refresh but the cube was the only object in the build script.
No other cubes or dimensions were maintained in the same build so I don't think it could have affected the use of TEMP tablespace.
Generally we don't use Atomic Refresh.
Thank you,
Vicky