on 02-03-2016 9:17 AM
Hi guys,
I have backend system on ERP 6.0 EHP6 with Central instance, one dialog instance and the database instance. Then I have a Portal with SSO configured with the backend. Number of users is considerably increasing and some heavy report-like transactions (Z reports) are causing some performance issues in the backend system. I have a few questions for clarification:
1) When a user makes a request in the Portal, the request is eventually fulfilled from which instance - the CI or the DI? In SM04, I can see plug-in users ...
2) In SM50, I see several RFC requests. Can this have an impact on performance of the CI ? What about the table TFRCQOUT? I can see it growing gradually. Is there some maintenance we need to do this table? I've seen note 375566 referring to this table.
3) I have some 500 users. My CI and DI have both 22 dialog work processes. At times, I can see in SM50 the number of free dialog processes reaches 0 (i.e all dialog WPs runnning). Is this sufficient?
I am planning to create a dialog instance to cater for the increasing load.
4) On my portal, I have 3 server nodes with 3GB heap size for each. Normally, I do not see Full GCs occuring as of now. Is there a way to know (t-code, report, etc) that the 3 server nodes are sufficient to handle the current load?
regards,
Suraj
Please check below points:
- Have you configured the load balancing in your ECC.
- Have you created the RFC connection from portal to ECC using load balancing.
- Check the database stats.
- Check the OS CPU, disk, memory utilization.
- Is any other activity going on on the DB server (ie backup etc)
- Is there any background jobs running which are taking too much resources.
Are the WPs are app server and CI are occupied? or only on CI.
As you mentioned if the Z reports are taking too much resources then I would suggest to speak to developer and request them to fine-tune. Also make some fields mandatory in the Z reports or sometime user training may be require on how to execute the reports efficiently.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Sunil,
RFC connection, server resources are ok. We had an issue with load balancing but it is now ok.
For the Z-Reports, we've started the fine-tuning, e.g fine-tune SQL, set mandatory fields and move reports to BI where possible. There were too many reports on ECC.
1) What is the capacity for a dialog WP - 1 WP can service approx 10 users...is this correct?
2) How about the portal performance (question 4 above)?
regards,
Suraj
Not all Z reports require fine tune, you need to findout what are the reports taking too much time and reason behind and finetune them first. Sometime end user training may be require to enter the values before executing the report. Also why they can't be scheduled as a batch job to run overnight if this is just for reporting purpose.
There is no hard and fast rule about WPs where 1 WP can server 10 users. If one user run long running program for an hour then that process will be blocked. The DIA work process should not be occupied for long time and should be available to serve other users.
Have you checked about the hardware disk/CPU/memory stats? Check with your infrastructure team about this. Is there any other programs than Z causing any issues? If possible can you please attach screen shot of SM66 when you have performance issue?
About portal issue, If backend is busy or not having free WPs to handle the requests then you will see the performance issue on portal as well. So, I would suggest you to fist concentrate on the ECC to resolve the issue.
What database you have in you landscape? have you modified any parameters or applied patches/kernel and then you seeing the performance issue?
Hi Sunil,
I understand your points. Indeed some reports are being run at night in background but we have too many of them on foreground during the day. The tables for some of the reports have grown very big.
Disk/CPU/Memory is ok. The reports are running for long in foreground (timeout is 5 mins) and so they are holding the dialog WPs for long.
In addition to this, we have SAP BusinessObjects 4.0 for reporting purposes and here also many of the reports fetch their data from ECC. Only a few of them fetch data from BI/BW. According to me, this is the main problem. We are progressively reviewing the reports. Also, we are planning for a second dialog instance, will this relieve the system?
I'll show the output of SM66 when I'll encounter a performance issue again.
Database is Oracle 11g and we have not done any patching or parameter change lately.
regards,
Suraj
Meantime please also check the oracle parameters as per SAP note 1431798 - Oracle 11.2.0: Database Parameter Settings
Hi Suraj
Review the below note, You need to tune your system , also provide the w/p log ,
88416 - Zero administration memory management for the ABAP server |
1431798 - Oracle 11.2.0: Database Parameter Settings |
http://scn.sap.com/docs/DOC-42062 |
1171650 - Automated Oracle DB parameter check |
1431800 - Oracle 11.2.0: Central Technical Note |
1591580 - Oracle 11.2.0: Oracle Home administration on Windows |
1431798 - Oracle 11.2.0 Database Parameter Settings Thanks Sadiq |
User | Count |
---|---|
93 | |
11 | |
10 | |
9 | |
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.