Skip to Content
author's profile photo
Former Member

Own Process type in Process Chain

Dear BI-colleagues,

The big international firm I work for has loads into BW 24/7. To avoid Master Data inconsistencies they want to implement the following loading strategy:

1. load Transaction data into PSA

2. load and activate Master data

3. load Transaction data from PSA into InfoProviders

This obviously to avoid SID-errors (MD check turned on) as transactions will be entered in R/3 while MD-load is running.

The problem is that the way process chains work is that the process chain wants the "infopackage to PSA" run in the same process chain that "pushes from PSA". Or the other way to describe it is that with process chains we cannot pick up data from PSA that was put there by an infopackage which was executed outside of the process chain. Thus we cannot de-couple them in order to run the master data process chains in between.

A solution we came up with is

- to create a custom Process type (copy of Process type "Read PSA and Update Data Target");

- Modify it to take some parameters like infopackage name, PSA table and data target;

- Add some logic that enables it to determine how many PSA requests (FIFO) should be pushed into the targets;

- Add some logic for unusual exceptions like if it does not find any requests.

Does anybody have hints/tips/sample source code etc?

(I already got hold of the How-to paper for creating custom Process Types).

Thanx a lot!

Regards,

Marco

Message was edited by: Marco Verbaan

Add comment
10|10000 characters needed characters exceeded

  • Follow
  • Get RSS Feed

3 Answers

  • author's profile photo
    Former Member
    Oct 02, 2005 at 01:42 PM

    Can't you simply:

    - load master data

    - load transaction data (with disabled errors)

    - load master data

    If transaction data is loaded and the master data isn't yet available a 'tentative' entry is made in the master data object. This master data will be filled later during the second master data load.

    Add comment
    10|10000 characters needed characters exceeded

    • Former Member

      What about this:

      - Load transaction data in ODS (buffer only)

      - Load master data

      - Distribute data from buffer-ODS to further data target.

      While filling the buffer you disable error-checking.

      In the second stage you can enable error-checking and implement further master data dependency.

      This strategy can easily be implemented with standard PC functionality.

      Message was edited by: T. Jans-Beken

  • author's profile photo
    Former Member
    Oct 03, 2005 at 05:34 AM

    Thanks for your input T, but tis will unfortunately increase overall loading time and that is unacceptable. The decision for the mentioned loading strategy has already been taken.

    Regards,

    Marco

    Add comment
    10|10000 characters needed characters exceeded

  • author's profile photo
    Former Member
    Oct 05, 2005 at 07:22 AM

    If you do create a custom process type, it might be simpler to create a "Wait" process type. You can put this in between the load into PSA and update from PSA into the target. The wait can be custom code which looks into some custom table which can be maintained by your conflicting process (the master data load in your case).

    Let me know if this helps and if u need further details

    Shailesh

    Add comment
    10|10000 characters needed characters exceeded