Measure Aggregated at Logical Level causes poor performance

Question

OBIEE 11g
We've recently implemented a new measure which involves aggregating a Period expenditure figure up to the Fiscal Year level.
I've duplicated the existing Period_Expenditure rename to FY_Expenditure, and then changed the logical level for the Time dimension to Fiscal Year.
I get the figures I expected, however the OBI SQL generated wasn't what I expected.
We now have the main SQL as before but now its left outer joining to a 2nd SQL statement that does the aggregation, which is pretty expensive in terms of Logical IO. Elapsed time going from 2 seconds to 20 seconds.
I would have expected the BI server to be clever enough to apply analytics to solve this problem - has anyone else had similar issues when using aggregates like this?
Thanks in advance,
Matt

Answer 1

I'm going to answer my own question here as I've managed to build a simplified solution, it may help others.
Basically the Analyses this is running for is:
select measures, FY_Measure
from fact_table
where period = 'MMYYYY';
Because we don't have the full set of data within this SQL to enable coverage for the FY_Measure (ie. we're at period level granularity) OBIEE is smart enough to create a 2nd query to get the data required for the year.
If we change the analyses to be:
select measures, FY_Measure
from fact_table
where FY = '2012';
It will quite happily create analytics to satisfy the measure - fairly straightforward really.