on 08-25-2010 6:48 PM
The documentation on data extraction (Master Data for Spend Performance Management) specifies that Invoice transactions are extracted from the table BSEG (General Ledger) . On the project I'm currently working the SAP ERP team is quite worried to run queries on BSEG as it is massive.
However extract files are called BSIK and BSAK of; which seems to suggest that the invoices are in reality extracted from those accounts payable tables.
Can someone clarify the tables really used, and if it's the BSIK/BSAK tables what fields are mapped?
Hi Jan
BSEG doesnt have the correct set of indexes that are needed for performant Spend data extraction. But at the same time BSEG has the complete set of data (as a cluster table it gathers the right set of data that Spend needs).
On the other hand BSIK and BSAK are indexed appropriately (they are infact index tables). But they dont store all info. So what the extractor starter kit does is that it figures out the Spend related data from BSIK and BSAK, cleared and un cleared invoices, and then goes to BSEG to actually get the data. So ultimately the data still comes from BSEG and there is no way around that.
I can understand the concern for extracting data from BSEG and BKPF. But after the initial load is done during implementation, the delta extracts, conducted on an ongoing basis, are not going to be that much of an hit to these tables.
Hope that helps.
Regards
Rajesh
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Jan,
Few additional mitigation thoughts which may help on the way as same concerns came up during our project .
1) Sandbox Stress testing
If available u2013 take advantage of an ECC Sandbox environment for extractor prototyping and performance impact analysis. BSEG can be huge (contains all financial movements), so e.g. BI folks typically do not fancy a re-init load for reasons outlined above. Ask basis to copy a full year of all relevant transactional data (normally FI & PO data) onto the sandbox and then run the SPM extractors for a full year extraction to get an idea about extraction system impact.
Even though system sizing and parameters may differ compared to your P-box you still should get a reasonable idea and direction about system impact.
2) In a second step you may then consider to break down the data extraction (Init/Full mode for your project) into 12 monthly extracts for the full year (this gives you 12 files from which you Init your SPM system with) with significant less system impact and more control (e.g. can be scheduled over night).
3) Business Scenario
You may consider to use the Vendor related movements in BSAK/BSIK instead the massive BSEG cluster table as starting tables (and fetch/lookup BSEG details only on a need base) for the extraction process (Index advantages were outlined above already).
Considering this we managed to extract Invoice data with reasonable source system impact.
Rgrds,
Markus
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Marked as unanswered because a follow-up question was added
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Jan
THere are no set guidelines.
As you have already guessed, it depepnds on Hardware, allocated memory, available processes, package size settings, and customizations made or removed (check documentation, some GL account related changes if removed can drastically speed up extraction, advised only if business requirements dont need that data).
Other BASIS + DB level tweaks and tunings would also have a big impact on performance.
Regards
Rajesh
Rajesh, a follow on question: is there any guidance or benchmark on what time it requires to extract invoice lines, say per million?
I assume it depends on the hardware as well.
can you share some experiences or indications?
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Thank you Rajesh, very helpful.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
User | Count |
---|---|
14 | |
4 | |
2 | |
1 | |
1 | |
1 | |
1 | |
1 | |
1 | |
1 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.