I’m waiting for the next CTP (September 05?) because it should work with cascading many-to-many dimensions. In a previous post I talked about the lack of functionality when you have more many-to-many dimensions chained together (I named this scenario “cascading many-to-many dimension”); I received feedback on this issue and it was a supported scenario, so what I described 3 months ago was a bug. What is important now is to understand performance in real-worlds scenario, because this kind of thing will tell us what dimensional modeling decisions will be best practice in different scenario.
This is a key feature to adopt Analysis Services 2005 in certain scenarios where today a SQL query could be better than a MDX one. The big challenge is that often many-to-many dimensions are paired with large dimensions (at least one). This bug prevented me to test some real-world scenario… If someone already has some “ideas” of numbers (we can’t perf-test a Beta, I know…) let me know!