OBI + Essbase filtering slow

Question

Hi,
I have a report that uses Essbase 9.3.1. as data source. When I create a filter on some dimension that includes members form more than one level, the report runs much slower.
Example:
we have a dimension "Organizational Unit" with levels from Gen1 to Gen6, the filter is defined for example as
Organizational Unit.Gen2,Organizational Unit = 'Country1' and Organizational Unit.Gen3,Organizational Unit = 'Region1'.
Alternatively, we can have a filter defined as
Organizational Unit.Gen3,Organizational Unit = 'Region1' which gives the same results as the above filter ('Region1' is a member at the next/child level of 'Country1'), but runs much faster.
Neither Organizational Unit.Gen2,Organizational Unit nor Organizational Unit.Gen3,Organizational Unit are present in the report itself (we only filter upon those columns).
This speed difference does not occur when both Organizational Unit.Gen2,Organizational Unit and Organizational Unit.Gen3,Organizational Unit columns are present in the report.
When examining the generated MDX query I have noticed there is a difference in the where clause of the query:
MDX1, both columns in filter:
+With
set [Organizational Unit5] as '{[Organizational Unit].[Country1]}'
set [Organizational Unit6] as 'Filter(Generate({[Organizational Unit5]}, Descendants([Organizational Unit].currentmember, [Organizational Unit].Generations(6),SELF), ALL), ([Organizational Unit].CurrentMember IS [Organizational Unit].[Region1]))'
set [Time3] as '{[Time].[MONTH_3_2010]}'
member [Organizational Unit].[Organizational UnitCustomGroup]as 'Sum([Organizational Unit6])', SOLVE_ORDER = AGGREGATION_SOLVEORDER
select
{ [Accounts].[EVA]
} on columns,
NON EMPTY {{[Time3]}} properties ANCESTOR_NAMES, GEN_NUMBER on rows
from [NISEPL.NISEPL]
*where ([Organizational Unit].[Organizational UnitCustomGroup])*+
MDX2, only 1 column in filter:
+With
set [Time3] as '{[Time].[MONTH_3_2010]}'
select
{ [Accounts].[EVA]
} on columns,
NON EMPTY {{[Time3]}} properties ANCESTOR_NAMES, GEN_NUMBER on rows
from [NISEPL.NISEPL]
*where ([Organizational Unit].[Region1])*+
The second MDX's where clause limits the unit "directly" to Region1, whilst the first makes a custom group of members and then filter upon that custom group. It seems this is causing the performance issue, which can be huge (up to 5x when working on large dimensions or multiple dimensions).
I don't have the option on limiting filters to just one, because the filter is being created through dashboard prompt on which we have all the levels of the dimension with constrain option so all levels get sent to the filter in Answers. Also, the report is to be filtered on different levels depending on the chosen prompt and constrain option is also required because of the size of the dimension.
Does anyone have some idea how to resolve this?
Perhaps creating the filter in a way that it would take/send to Answers report only the last chosen level on the prompt?

Answer 1

Thanks John that worked. But when I applied the patches again I am getting the same Error, " Unable to open the application".
I cleared the Temporary java files again, even then i am getting this error.
Edited by: 863529 on Aug 31, 2012 5:41 AM
Edited by: 863529 on Aug 31, 2012 5:43 AM