Performance Impact on SAP Source system

[Former Member]

Hi All.

I am busy with a project profiling SAP ERP source data, using SAP Information Steward 4.2, SAP Data Services 4.2 and Information Platform 4.1. We are currently working in the DEV environment.

I would like to know what the performance impact will be on the SAP ERP system, if we had to apply profiling in the Production environment.

I have gone through the Information Steward 4.2 documents and it does NOT mention anything regarding performance impact on the source system when conducting profiling directly on a source system such as SAP ERP. Does anyone have SAP official information on this?

Would really appreciate your help in this regard.

Thanks.

Regards,

Thatoyame Lolwane

Email:thatoyame@gmail.com

[Former Member]

We haven't looked specifically into profiling, but we always keep the BASIS team in the loop and sit with them to monitor the impact on the dev and prod during several full-size test runs.  We've run significant volumes of data through validation rule tasks directly against production systems and never had any issues with impacting the performance of the ERP system.  I would suggest you work closely with BASIS and use the monitoring tools available on their side. 

[former_member226511]

Thatoyame,

I would suggest reviewing the SAP Supplement Guide for Data Services.  In there it talks about performance impacts (among other considerations) for the various methods of extracting data.  InfoSteward effectively creates a Data Services dataflow, so the considerations apply for both applications.

http://help.sap.com/businessobject/product_guides/sbods42/en/ds_42_sap_en.pdf

[former_member187605]

Best practice is to copy the necessary data from ECC to a dedicated staging database first. Then do all your profiling in staging. This will minimise impact on the source system.

Implement the copy process with DS. You can make it an incremental job, if you want your profiling activity to be repeated at regular intervals.

[former_member186454]

I would not call it best practice especially when working with sensitive information. The key differentiator is that IS can profile the data without having to extract it to another system. IS uses system information and functions to do this and the impact in most instances is therefore much less than doing a full extract of all the data.

[former_member187605]

You're absolutely right if you profile your data only once. But unfortunately, data profiling often is an iterative process. And a typical Basis team does not want you to repeatedly run large extracts on the ECC system.
FYI: IS does not access anything at all. All it does is generating DS jobs that do the hard work. If you profile an ECC table every week, the DS job generated will access the full ECC table every week.  If you stage your data properly, DS will only load the necessary data the very first time and do an incremental load afterwards. No need to access the full source table over and over again, hence lower impact on ECC.