SSAS introduced the concept of many-to-many relationships between dimensions, an outstanding development that partially overcomes the rigidity of strict star/snowflake schemas. But with this feature came all sorts of warnings from Microsoft.
Entries in OLAP Design (4)
It's rare that a cube will contain transaction details. So once a user drills down to the bottom of a cube, it's common to want to see the transactions that comprise a given value. That's what drillthrough is for, or at least that was the case in the old days.
I receive a lot of technical questions via email, often involving complex topics. Unless your question is general in nature, it's unlikely I can provide useful feedback unless you frame the question in AdventureWorks context, especially if it involves MDX.
I'm glad to field questions, but please be patient if I don't reply right away.
Add June 1, 2007: This 2003 article is still pertinent with Analysis Services 2005.
Rare is the OLAP cube that does not include a so-called Time dimension – a dimension that allows data to be analyzed over time. Though often named simply Time, the typical Time dimension is actually a date dimension, with a hierarchy consisting of Year-Quarter-Month-Day or the like. Despite the unique ever-presence of Time, this is the one dimension that developers who are new to SQL Server Analysis Services often get wrong. This article will explain the correct way to implement Time dimensions in Analysis Services, and the benefits thereof.
The full article is available at SQLJunkies.com.