So I got to thinking, and after searching around a bit, I didn’t find much, “A blog post is in order!” So here is what I was researching.
Where and when is it a good idea to start breaking out to multiple cubes?
What is the best practice about splitting up cubes or segmenting data into different cubes?
What are some of the reasonable limitations of Microsoft SQL Server Analysis Services?
…as I dig more, a write up will follow!
We had this question in the past. Microsoft’s guys recommendation is to fill as much data as you can into one cube. In SSAS 2005 one cube can have many facts, so in terms of SSAS 2000 every cube is a virtual cube (in a way). So, as long as you can, do not split.
I never checked out the exact limitations of SSAS, but I can sure you that they’re very far to 90% of the market. Look at Barnles & Nobles that use SSAS 2005: They know how to manage their DB so the performance is maximized. For exmaple, they split the cubes into many partitions according to the date.