What's the usage of the associated columns?

Question

In business model, logical columns are associated to a level of dimension. But the columns are nether primary key nor used for display. What's the purpose of these associated columns?

Answer 1

To flesh out our example a little more.
You have a Day Dim, and a Month Dim table. Your month name logical column is mapped to both tables via different Logical table sources. Your content levels are set accordingly for those LTS's.
You have a fact table at grain Day and an aggregated fact table at grain Month. You have a measure (dollars, quantity, whatever) mapped to both physical tables through two LTS's with content levels set (like the Time Dimension).
So when you query Month Name and Dollars, the BI Server uses the aggregate table as it can serve up both the dimension attribute and the measure from the aggregate table. It knows its more efficient because the ratio of the number of elements you defined in the hierarchy tells it that there is approx 30 times less data in the month aggregate, by the ratio of 12 months to 365 days.
So far so good.
Lets introduce more dimensions, a.n.other dimension is joined to the base fact table but is NOT in the Month aggregate. You model it all as per normal, when you drop this extra dimension into your report, the BI Server recognises that there are no aggregate sources that can give the correct answer across all required fields in the report so drops to the next best fit at the grain requested in the report, in this simple case it would default back to the base fact table as that is the only fact table capable of serving up the month and any attributes from the other dimension. In more complex models you might have a few aggregate tables with both higher grain (higher levels) or less dimensionality (less rows). The existance of physical layer joins tell the BI Server what can be joined for the correct dimensionality, the setting of LTS levels tells it which aggregate tables would be more efficient.
Dont get confused with setting a level on a measure - thats something different (level based totals, % of Year etc. etc)
Got it?
These might help :
http://gerardnico.com/wiki/dat/obiee/fragmentation_level_based
http://www.rittmanmead.com/2006/11/aggregate-navigation-using-oracle-bi-server/
http://obiee101.blogspot.com/2008/11/obiee-making-it-aggregate-aware.html
let us know if any more Q's.