Splitting Dimensions on Attributes for reporting

Question

Does anyone have experience or a good idea on how to most efficiently allow splitting a dimension on attributes for reporting custom cells. We are working in 11G. First let me explain
We have a dimension Brand that has a level based hierarchy.
Total
Category
Brand
Item
The Item level has many attributes, of course it has Category and Brand but it also has other Attributes that apply only to the Item level, we will say Flavor and Size for example. The user might want to see a report with Flavors on Columns and Size on rows, or Flavor on columns and Categories on rows.
We have come up with 4 ways to do this, 3 on the fly and 1 predefined.
1. We use the BRAND_VIEW view to limit to the chosen attributes and the Item level then join to the CUBE_VIEW and sum BRAND. This only works for measures with all SUM additively
2. Use the Java API and Source objects to limit BRAND to the ITEMS that meet our criteria, create a custom member in brand and report on it. This gets around the additively problem but is currently not an option til Oracle patches the 11G API to allow session only member creation.
3. Write a DML program to limit the BRANDS and then report on the TOTAL. Again this only works for SUM.
4. For the above example we would build 3 dimensions for BRAND instead of 1. We would build the BRAND dimension as defined and a FLAVOR and SIZE dimension. This increased the size of the build and fact data, also greatly increases the build time. Also this forces the User to see many more dimensions then they are use to seeing.
If anyone has dealt with an issue like this I would greatly appreciate hearing how you handled it.

Answer 1

Clyde, how are you sir? I was curious on this too - so I took a look into it. Seems that the as long as the attributes are marked navigational they come along into the BIA indexes. See excerpt and link below.
[http://help.sap.com/saphelp_nw70ehp1/helpdata/en/43/bcf9bfdd052ec5e10000000a11466f/content.htm|http://help.sap.com/saphelp_nw70ehp1/helpdata/en/43/bcf9bfdd052ec5e10000000a11466f/content.htm]
Step 1: Creating a BI accelerator index
When you execute this step, the system creates the indexes for the tables of the InfoCube star schema on the BI accelerator server, as long as they have not already been created by other BI accelerator indexes. These tables consist of the fact and dimension tables of the InfoCube as well as the master data tables that contain the required SIDs, the S, X, and Y tables of the InfoObjects
X = time independent navigational attributes and Y = time dependent navigational attributes, so it looks like these are stored too.
Hope this helps!
Justin