|SQL Server 2008: Attribute Relationship Designer|
|Written by Stacia Misner|
|Monday, 29 October 2007 15:33|
The dimension editor in SQL Server 2008 Analysis Services (in the July CTP) is supposed to improve the development experience by making it easier to work with attribute relationships. I read a comment from Vladimir Chtepa on Mosha's blog that no UI improvement was needed for attribute relationship design, but from my experience this person's opinion is not the norm. At least not in my classrooms. The people who generally have the hardest time understanding the arrangement of attribute relationships in SQL Server 2005 Analysis Services were those who had experience with Analysis Services 2000 (AS2K). Compared to defining levels in AS2K, which displays the relationship between levels in a top-down approach withhigher levels on top, the attribute relationships in Analysis Services 2005 (Yukon) seem upside-down by taking a bottom-up approach. That is, you connect a "higher-level" attribute to a "lower-level" attribute. For example, in a time dimension, in AS2K, you define Year-Quarter-Month-Day in that order from top to bottom whereas in Yukon you add Month as an attribute relationship to Day, add Quarter as an attribute relationship to Month, and Year as an attribute relationship to Quarter. My students who started their Analysis Services experience with Yukon had less trouble with this concept, but still commented that it seemed backwards and required some extra thought. I think the problem is related to how people visualize hierarchical information and the Yukon UI does not lend itself well to visualization of attribute relationships particularly if there are a lot of attributes in a dimension and the related attributes are spread far apart from one another. Therefore, I think the new attribute relationship designer will help a lot.