I have a very simple scenario which is causing me an amazing degree of difficulty and I hope you guys can help.
I am running SSAS 2008 R2 and have a Cube with 1 Measure Group with a materialised measure set to a sum aggregate type and 1 related dimension. The Fact Table originally had 17 million records and the Dimension has 21000 members. The Dimension has a hierarchy which has approximately 7 levels and the lowest level has 21000 unique values (the Product Code). The Dimension also has individual attributes (outside this hierarchy) which allows the user to select these attributes individually. There are no attribute relationships defined in this case (have removed them for troubleshooting purposes).
In Excel, if I use the hierarchy, the whole thing can be expanded out in a matter of a few seconds but if I build the same levels of information using the individual attributes, the final attribute (21000 unique members) will take forever and often result in the Excel client crashing or running out of resources. If the attribute is used on it's own and against a dingle measure, the result displays in seconds.
This scenario has been simplified but demonstrates the issues.
Here is what I have tried so far and with the same results:
- deployed to SQL Server 2012
- created attribute relationships
- used a different attribute with the same level of cardinality
- reduced the Fact table to 1 million records
- run Excel from the client and the local server
The server has ample resources.
I am about to embark on Profiler but not quite sure on what I am looking for. 21000 tuples and the intersection with these preceding attributes doesn't seem like a particularly onerous task. Can anyone please advise on what this could be and what I should be tweaking.
Thanks in advance.