How does SID generation on DSO level affect InfoCube load?

[#1]

Hi experts,

I am having big performance issue with data load to InfoCube. In the DTP monitor, the step "Conversion of Characteristic Values to SID" takes about 7-8 mins for each data package of 50000 records. I need to load about 20 millions data & this would take ages.

Below illustrates part of the data flow in my scenario:

DSO A   
DSO B   ======>  DSO X =====> Cube 
DSO C   

In the entry layer, I have DSO A, B, and C which contain header, line item records as well as master data. The option "Generate SID upon Activation" is checked for DSO A, B, & C.

Records from these 3 DSOs are consolidated & enriched into DSO X. In order to reduce load time I have unchecked "Generate SID upon Activation" for DSO X.

Data from DSO X will be written to the Cube 1:1 without any additional transformation logic.

My questions:

1. Since I have generated SIDs for the chars in entry layer, why it still takes so much time to for the SID conversion during Cube load?

2. In this case, is it necessary to generated SIDs in DSO X?

Thanks in advance for any advice. I am in desperate need for a solution to this load problem

Regards,

Meng

[anindya_bose]

Hi Joon

Some questions..

1. Do you delete and reload this cube every time? If yes, do you delete Dimension Ids as well when delete the cube data ? [ Check in Process Chain settings]

Deleting dimension table before cube loading increase the time in in step "Conversion of Characteristic Values to SID" as to create dimension ids system needs to read all the SIDs and with a combination of SIDs dimension ids get generated.

2. Are your Cube dimensions more than 10% of the Fact Table? Or do you have very big dimension table? Please run report SAP_INFOCUBE_DESIGNS to check if you have problem with dimension size. System will show 'bad' dimension in red for a particular cube

You can also think of Number Range Buffering so create DIM and SIDs faster.

Generating and not generating SIDs for your DSO X should not have any impact on the DTP load to cube.

Regards

Anindya

Edited by: Anindya Bose on Aug 30, 2011 3:01 AM