Low performance with more of than 25 dimensions

[Patricia]

Hi Sir,

I've worked with more of than 25 dimensions, it's have had problems with performance. I've perceived that the time to build the PivotCube (using active:= false and active:= true) with 1600 records of the TClientDataSet is same with 7000 records. But I did test: I leave only 10 active dimensions and it better performance with same numbers of records.
what I can do possibly?
Please, some suggests?
I thought to use TPVDimToolBar with inactive dimensions, but I don't obtain sucess because this component list only active dimensions (enable = true).

I would like to ask excuses for my english
Thank's
Patrícia

[AlexZencovich]

How much is that 'low'? 100 milliseconds or less? By my tests with PivotCubeBenchmark it is

For 32 dimension and 1,700 records cube built take 2.5 seconds
For 16 dimension and 3,400 records - 1.6 seconds
For 10 dimension and 10,000 records - 1.6 seconds

but

For 16 dimension and 100,000 records cube built take only 7.7 seconds.
For 32 dimension and 100,000 records cube built take only 11,3 seconds.

(PivotCubeBenchmark is available with trial package, anybody can test on own PC)

You should expect - if you have increase number of dimensions/measures/records it mean more memory occupied by cube so it may use increase build speed for some little time.