Essbase 7.1 server performance tuning

Question

HI Friends,
I have a cube with nine dimensions,with block storage.
The lowest level in time dimension is Month.
When I load this cube with 3 yrs of data, it is taking taking 20hrs to rollup.
The dataload takes 2hrs.
There are no complex formulas in this cube, it is a simple rollup.
I have index cache at 500mb and Datacache at 700 mb.
I have only 2cpus on this box.
when the cube calculate the size of the cube is 180GB
Measures     Dense          13
Time     Dense          33
Gender     Dense          5
Ethnicity     Sparse          10
EmplClass     Sparse          26
Workgrp     Sparse          74
RefSource     Sparse          23
JobGrp     Sparse          89
Department     Sparse     20444
Please suggest me the right congigaration to make the calculation faster.
I would appreciate any ones help.

Answer 1

I would begin by removing any current dynamically calculated members in the sparse dimensions and make the rollups for Ethnicity and EmplClass all Dynamically Calculated. Then I would assume the following base calc (based on a "calc all", the actual calc would depend on any deviation from that):
SET MSG SUMMARY;
SET NOTICE LOW;
SET CACHE HIGH;
SET UPDATECALC OFF;
SET CLEARUPDATESTATUS OFF;
SET AGGMISSG ON;
// Optional: CLEARBLOCK UPPER;
CALC DIM( Measures, Time, Gender, Workgrp, RefSource, JobGrp, Department);If (for instance) you fix on time, the fix would require removing the Time dimension from the calc dim list.
I would also ensure you are using buffered I/O and have a decent allocation of memory to the calc cashe (SET CLALCCACHEHIGH setting in the Essbase.cfg file).
With this as a baseline, I would then try other combinations. The problem is that these combinations would require knowing the number of calculated members in each dimension (non-leaf, aggregated, or formula based members). The use of shared members in alternate hierarchies also has an impact on what I would recommend, as well.
However, the baseline suggested above is the most likely cost/benefit leader if the cube is a straightforward rollup with no "complications" in the Ethnicity or EmplClass dimensions.