on 11-30-2011 6:55 PM
Dear all ,
We are doing a scenarion from JDBC to IDOC Async and also IDOC to JDBC Async. The JDBC to IDOC scenario we are expecting to process 6 million messages at the end of each month. This needs to be done in a span of 4 hours.
Is PI 7.1 capable of handling such huge volumes in a short span of time? The size of XML messages are between 1 mb to 4MB.
Please advice . Is it a good idea to add additional app server or Increase RAM?
In sap market place there is siazing only for PI 7.3,. I also checked the webinar for Quicksizing SAP Netweaver PI 7.1 but need input from experts.
Thanks,
Teresa
Hi,
do not focuse you only on PI, and not only on hardware. your interface development should be also efficient.
Check also with the database if there is a timeout limit for the JDBC connexion.
Check with ECC, because depending of which inbound IDoc you use, the delay could be long for SAP to do all controls (like in a PO) for all your 6 millions.
Be carefull also with the connexion PI -> ECC, because the number of parallel RFC connection is limited (I think).
You probably will not be able to use "maxOccurs=Unbounded" with only one source message, because that means ECC receives only one Message that it has to cut in 6 millions of idoc: you could have ecc issue to store/write idocs in SAP table. Some customize have to do.
Check if you can parallelize treatment from end-to-end. be carful of number of authorized parrallel connexions (ecc and database).
For such a volumetry done once a month (not alone, but with other interfaces...), I will strongly recommend you to involve SAP to your discussion, for the hardware, the SAP custo to do, and why not also for the development methodology.
regards
Mickael
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Thanks Mickael for the quick response. You are correct about the number of RFC connections created in Parallel.
We currently had an issue where all the jobs were running in parallel to post the idocs and took all of the system resources.
What kind of customization are you refereing too on the SAP side?
Do we create a message with SAP to ask them about hardware requirement for such a high volumetry data?
Thanks,
Teresa
Hi,
>> What kind of customization are you refereing too on the SAP side?
I need to check if I referenced that somewhere. but not this evening, tomorrow (france).
>> Do we create a message with SAP to ask them about hardware requirement for such a high volumetry data?
you can try... but I fear the answer should be: your PMO should contact your SAP commercial contact for consultings!
Mickael
Hi Teresa,
to know which parameter to customize, see these oss notes:
Note 74141 - Resource Management for tRFC and aRFC
Note 1375656 - PI_SystemParameters_v03 (Default values)
You can run program RSPFPAR to know all parameter values that you have.
See also this old doc (sorry)
[Scaling High-Volume Messages with SAP Exchange Infrastructure|http://www.sdn.sap.com/irj/scn/index?rid=/library/uuid/c02b4fc0-be11-2a10-4984-9713585ca128], but by searching in SDN or service.market place, I quite sure you will find this one for PI 7.1.
regards
Mickael
> Is PI 7.1 capable of handling such huge volumes in a short span of time? The size of XML messages are between 1 mb to 4MB.
PI 7.1 supports message packaging. This scenario is right candidate to try and your size of the message is pretty much acceptable for message packaging. Follow this link for configuring and use it. Thoroughput and performance is considerably improved.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Baskar,
Thanks for the link . For the JDBC to IDOC scnario [SQL to SAP] PI 7.1 picks up all the reocrds from teh sql table and in SXMB_MONI i see one inbound message with all the XML payloads for the different accounts and line numbers.
I see the problem in SAP where once the idocs are created in SAP it takes a long time to post in SAP . It generates so many parallel RFC connections and kind of puts all other jobs to hold.
Will message packaging help in the above scenario?
Thanks,
Teresa
Message packaging assigns work process for every individual pipeline step. So if we have multiple messages from sender to target system it will be very helpful. Thoroughput and performance are very good. we have in our syteem. If you create a single target message (for all the messages) and let adapter engine to split messages after processing pipeline step, this might not be fully effective.
User | Count |
---|---|
94 | |
11 | |
11 | |
10 | |
9 | |
7 | |
6 | |
5 | |
4 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.