TM1 Forum

New to Cognos!!! but been working on Microsoft Analysis Services,
During processing of cube in SSAS aggregations are preprocessed and stored on disk and it does not process all possible aggregations but based on DSV and cube design and hierarchies and attribute relationships design time aggregations are created.

But my customer in cognos says that in Cognos all possible aggregations are created like if I have 10 dimension with each having 10 attributes.. Cognos creates aggregations for all possible combinations of attributes across dimension and in dimension as well , which is not practically possible.

Based on this I have 2 questions
1. Is the understanding of my customer correct?
2. If Cognos is creating aggregations is there tool that would let me look at these aggregations. (as they claim they do not need to create aggregations or look at aggregation design as all possible aggregations are created by Cognos)

It depends what you mean by "Cognos" if you are talking about "Cognos TM1" then this is correct in practice in how it may appear to a user but not technically correct in terms of how the aggregations occur.

TM1 cubes are based entirely in RAM, there is no pre-aggregation whatsoever and aggregations are calculated on demand when requested by users in a cube view. TM1 also does not store any data for zero intersections, therefore null cube space effectively does not exist.

This means that unlike disc based OLAP technologies TM1 does not suffer from "database explosion" and it is possible to have highly sparse cubes which large numbers of dimensions with very little performance impact in terms of calculation speed and response time.

Therefore TM1 can deal with returning any combination of dimension intersections upon request (but it does this with no pre-aggregation). This may be what your customer is talking about.

If your user is using Cognos Enterprise Planning then it depends on which element they are using. If they are using contributer then thing become a little different as it is completly relational. If however they are using analyst then it is true. In analyst all possibilities are calculated regardless of whether they contain data or not. Add this to the fact that all data calculation are local machine based and you hit the recommended cube size limit of around 30 million possibilites. This is why all analyst models end up in lots of small cubes linked together.

I hope that helps,

Jim.

If you're MSAS, you may also be looking at Powerplay Cubes.

Unfortunately, I don't know the answer to the question in relation to PP cubes.