Comments on: Redundant Attribute Relationship Health Warning!

By: Chris

Chris — Tue, 27 Jun 2006 08:42:03 +0000

I think people on the AS dev team would be happy to admit that BOL wasn\’t where it should have been at RTM, but I do know they have been working hard on it ever since and that it\’s getting better.

Re AS2005 books, I\’ve already done one or two and still have \’The Microsoft Data Warehouse Toolkit\’ to go. If any authors out there want to send me free copies of books I haven\’t got to review I\’d be more than happy to receive them

By: Unknown

Unknown — Mon, 26 Jun 2006 02:14:34 +0000

… "this is something everyone designing AS2005 cubes needs to be aware of" …

If it was something everyone should know, don\’t you think they would have put it in the very extensive and helpful books online for AS2005? I continually find myself wasting valuable time trying to find good information in books online.

More to the particular topic, I was blown away by the problems that redundant attribute relationships cause in AS2005. When I read the new warning, I initially thought to disregard it because it only spoke of problems when the dimension was being used on measuregroups with a granularity that was different than the key attribute. I think they understate the potential problems.

Also, I find it hard to believe that "less is more". Can\’t SSAS just disregard the extra meta-data information that it finds to be extraneous? I have always believed that the more information you have available to you, the better decisions you can make when solving software problems. Attribute relationships should be no different.

Anyway, thank you for your blog. It\’s time better spent than reading books online. Would you mind reviewing some new AS2005 books sometime?

P.S. I checked Adventureworks and it has redundant attribute relationships too.

By: Chris

Chris — Mon, 27 Mar 2006 10:21:18 +0000

I couldn\’t have put it better myself, Jon. Thanks!

By: Jon

Jon — Mon, 27 Mar 2006 09:03:50 +0000

Rob,

Chris isn\’t suggesting the wizard creates two redundant relationships for you, he\’s saying that the wizard will automatically infer that every non-key attribute is related to the key attribute and will then create these relationships automatically (i.e. City->State and City->Country in his example). Since this will typically not reflect the complete set of relationships in the dimension (i.e. State-Country), the user must then manually add in relationships between non-key attributes, and remember to remove the extraneous relationships to the key attribute (i.e. City->Country) so as to avoid the problems with redundant relationships.

The exception here (as you imply) is that if you have a snowflaked dimension schema then the wizard will infer the "correct" relationships, based on the table keys.

Jon

By: Rob

Rob — Fri, 24 Mar 2006 23:37:14 +0000

The wizard will *not* create two relationships for you. The wizard will only create relationship to dimension/table keys.Theere are numerous problems with redundant attributes, We\’ve added a check in the dimension designer for SP1 to detect and report this condition.-rob

By: Unknown

Unknown — Fri, 24 Mar 2006 11:52:55 +0000

Hey, stop giving other people the credit for my amusing and insightful comments, Chris!

Colin