IBM Cognos User Manual page 196

Adjust the Auto-Partitioning Parameters
You can use the Auto-Partition tab on the PowerCube property sheet for each cube
in your model to test different settings and see which ones yield the best results
for your situation.
Before you begin
Before you begin, ensure that any previously defined manual partitions are cleared,
and that the optimization setting on the Processing tab of the PowerCube property
sheet is set to Auto-partition.
Procedure
1. Click the Auto-Partition tab and experiment with various test settings, as
2. Click the Dimensions tab, select the dimensions that you want to include in
3. Click PowerCube Partition Status on the Tools menu, learn how the automatic
Partition Manually (if Required)
If your model design does not support auto-partitioning, or if you have a model
with an unusual structure, you may achieve better performance by setting up the
partitions manually.
182
IBM Cognos Transformer Version 10.1.1: User Guide
follows:
v Specify an Estimated number of consolidated records. Consolidation
combines records that contain identical non-measure values, reducing cube
size and improving run-time performance. Start with the default. Later, you
can check the log file to see the actual number.
v Adjust the slider in the direction of Faster cube creation or Faster cube
access, as appropriate for your current requirements. By default, the slider is
centered.
v Specify a Desired partition size, which is based on the consolidated record
count. The default number of categories is 500,000 but you can reduce this
number in incremental steps, testing various settings until you achieve
acceptable run-time performance. When you build your cube, Cognos
Transformer automatically groups all categories into partitions of this size, in
a way that is optimal for your model. Partitioning stops when the number of
records in the summary partition drops below this specified partition size.
v Specify the Maximum number of passes, beginning with either five (5), or
one (1) pass for every manually defined partition level. The partitioning
process continues, one pass per level, until this number is reached, or until
subsequent consolidation passes fail to yield a further reduction in the
summary partition size.
Tip: Set this value to the minimum number of passes needed. If tests show
that the last few passes did not significantly improve performance, decrease
the number to achieve an appropriate balance between data access and cube
build times.
your cube, and click OK.
partitioning algorithm has distributed your records during your partitioning
tests, and act on the results. For example, if both your summary partition and
your level-one partition have the same number of records, then the summary
partition has not been consolidated. You need to increase the number of passes.
Tip: You may want to perform some manual partitioning tests, comparing
those results with the performance gained by auto-partitioning and choosing
the best solution for your particular OLAP reporting situation.