When it comes to tuning performance in HPCM it shows that HCPM’s Essbase Databases do not respond to the traditional hourglass outline tuning methods.
This is due to the fact that HPCM’s calculations are not the traditional dimension aggregations that respond well to the hourglass approach.
Instead HPCM is performing many leaf to leaf calculations that are sensitive to a very different set of tuning changes.
The tuning choices are usually unique to the individual model but there are some common characteristics that impact performance:
Dense / Sparse dimension combinations in the stages.
The nature / type of assignments made between stages.
The density and size of the data in each stage.
The Essbase database cache settings.
The HPCM JVM memory allocation settings.
Tuning the performance can show improvements in the range of ‘hours to minutes’.
In order to assist with the tuning the attached spreadsheet has been created. This spreadsheet can be used for collecting information about model characteristics which will enable Oracle Support to assist.
In addition to data captured in this file there are model queries that can be execute in SQL that will provide information about the use of rules and drivers (see tab 3 in the attached spreadsheet).