As I explained in the EE forum (http://forums.pentaho.org/showthread.php?t=68902) it is not possible to restrict the scan for Aggregation Tables to a specific schema.
I'm working a lot with Oracle DB's with many different Schema's with hundreds of tables/views in there. Mondrian's scan through the data dictionary to find Agg Tables takes a lot of time (couple of mins at least). The scan for agg tables happen everytime a MDX query is executed, so this is an absolute critical blocker for using agg tables in an large Oracle DWH.
Rolf has proposed a change in the API (
MONDRIAN-511). I can see the point of not doing the change in there. Julian suggested a new Schema attribute for AggName, which sounds quite reasonable. However, AggName definitions would add a lot of verbose lines to the Cube definition and is not as flexible as defining an AggPattern. With only a few lines you can easily do a transparent definition of the agg tables.
So I propose adding a Schema Attribute to AggPattern and optionally to AggName (thats used by PAD, so I think it would be good to have that in there as well)