SAP system performance problems

[Former Member]

Hello,

I´m trying to understand something about how and what can be done to improve our SAP system performance.

First of all this is a SAP ERP system that is installed in a windows server 2003 on a MS SQL Server database (in a cluster - cluster administration). We noticed that the performance of this SAP system is not the better and some users told us that some actions during the transactions are slow. The two servers (two cluster nodes) have a lot of RAM memory and also page file so I believe this problems are not related with this particulary things.

Can you told/teach me how can I check and measure if the system are with some problems!? Which transactions I must check to make this analysis and how can I correct/improve this situation?

Can you tell me please some tips regarding this type of problems?

Thank you,

samid raif

[manumohandas82]

Hi ,

From the screenshot dont think that you have any sort of memory issues  due to the parameter setting

of the memory .

Please provide a screenshot of the TXN ST06 .

You have mentioned that users are facing performance issues when executing certain transations . Have you done a trace for these transactions /users using ST12 or ST05

Please Dig in more using transaction ST03N / STAD  . You will get more details

Response Times: Rules of Thumb - CCMS Monitoring - SAP Library

All performance problems may not be memory . It could be CPU , network etc .

Thanks ,

Manu

[Former Member]

Hi Samid,

From the screen shot of ST02, its clear that the export/import buffer overflow has occured and high number of swaps has occured. High number of swaps will always a major concern in performance point of view.

Please refer the below notes to improve the export/import buffer

373986 - Overflow of the export/import buffer

702728 - Profile parameters for export/import buffer instances

649613 - Export/import buffer SWAPS

Regards,

Nikhil

[magexposito]

Thank you Nikhill, I completely miss the Export/Import Buffer hit ratio. It is fine if this buffer has Hit Ratio % above 80%, but in this case the Hit Ratio is below 80% and the buffer size should be raised.

Anyway, in this case I think that the system performance issue is not related with a wrong seize on the buffers.

Best regards.

[magexposito]

Hello SS,

A performance problem on the system can be caused for different reasons:

- Not enough resources during the process. For example, the extended memory of the system is exhausted an the process finish with an error. Or the CPU of the server is not enough and it will cause a bottleneck.

- Network problems or communication problems, between the different elements of the SAP systems or between users and system.

- Bad use of transaction by end users or user appreciation of the system performance.

- Etc.

I will check the following things to find where the performance issue is:

- Check the available resources of the system: Is there enough memory (roll, extended, heap) in the system (ST02)? Is the CPU at 100% or near 100% (ST06)? Is there enough physical memory all the time? What is the hit ratio % of each buffer on transaction ST02?

- Is there any dump on transaction ST22 that are related with insufficient memory on the system? Is there any message related with performance issues on transaction SM21, like for example connections timeout, message related with not enough resources, etc?

- Check the response times  for instance and logon groups on transaction SMLG. Which is the value?

- Check the hourly response times on transaction ST03. Compare them with previos values from other days, weeks or months, are they higher?

- Is the performance on the database good? Check transaction DB02. How is the CPU use on the database server?

- All the users complains about system performance or just some users? All the transactions or actions done in the system are affected by the performance problem or just some transactions?

Just to know, SAP doesn't make any real difference between Physical and SWAP memory. What I mean is that you define a pool of Virtual Memory between Physical and SWAP memory, then assign it to the SAP system with the system parameters and the SAP system uses that pool as it wants, so it is a normal situation if you find that SAP system is not using physical memory even if there is memory available. Of course there can be a problem related with wrong memory settings (for example, the amount of swap space you defined considering the physical memory of your server) but SAP systems doesn't always use physical memory.

Regards.

[Former Member]

Hello Miguel,

Thank you very much for the information detailed/explanation!

But can you check my previous message here, the values in ST02 transaction? Have you have an opinion or any tip about that values?!

Thank you,

samid raif

[magexposito]

Hello Samid,

The values showed on your screenshot are good:

- The Hit Ratio % of each buffer are between the recommended levels by SAP.

- There is no swaps on the buffers, just the Export/Import buffer has swaps and it is acceptable.

- There is enough roll, extended, paging and heap memory and it is not completed anytime.

Just one thing: You said on your last post that your SAP server has about 24GB RAM memory. Your extended memory is 13GB size and it should be higher because the virtual memory on your server should be 24GB RAM + Swap space defined (3xRAM Memory at least). Anyway, this is just a recommendation and it is not the performance issue.

Check the rest of topics I wrote on my answer. Also, how many work process has your server? Is there always dialog work process available all the time?

Best regards.

[Former Member]

I'm wondering if extended memory is set sufficiently - what is the value for EM/TOTAL_SIZE_MB?

In ST02, what are the figures shown for extended memory?

[Former Member]

Hello Ross Armstrong,

I don´t have any parameter with that name EM/TOTAL_SIZE_MB! This is a SAP ERP 4.7 version... I don´t know if it's related with this version or not, but I can't find that parameter.

In ST02 I see the following values:

Any tip?

Thank you,

samid raif

[Former Member]

Hi Samid,

Please check the below URL for the performance tuning:

ABAP
Performance and Tuning - ABAP Development - SCN Wiki

   SAP R/3 Performance Monitoring and
Tuning

http://scn.sap.com/docs/DOC-47062

Regards,

Rishi

[Former Member]

Hello to all,

First of all thank you for the information and the knowlegde that you share!

I will try to explain you what is the main problema here, that is very strange thing (or not... I don´t know exactly ).

As I told you this is a cluster platform, the node 1 has the MS SQL Server database and the node 2 has the SAP instance. Both servers have 24Gb of RAM each and the problem is the node 2 (where the SAP instance is) only use 3-5 Gb of RAM memory and it begins to catch the page file memory! I don´t understand why... because the PHYS_MEMSIZE has the value 20000 (20Gb) and I believe the server must use more RAM memory instead using the page file with a lot of free RAM memory!

Any idea why this is happen?

Thank you,

samid raif

[Sriram2009]

Hi Samid

1. On both nodes increase the virtual memory=physical ram(24gb) x 3 times = 72 GB

2. On both nodes most of the parameters should be same(Like profile parameters No of workprocess,Memory buffers values, physical node configurations) because at any point of time one of the node will act as a CI,

3. After rebooted the system initially it takes 3 to 5 gb, based on over all usage it will increase  using the physical & virtual memory.

BR

SS

[Former Member]

Hi SS,

Thank you for your answer Sriram.

1. Regarding that... we already increase the virtual memory to 72Gb

2. The instance profile has the following parameters:

I also checked the ST04 transaction:

Any tips?

Thank you,

samid raif

[Sriram2009]

Hi Samid

1. You can keep referenece of the SAP Note to increase or decrease  the memory & buffer values and also keep the memory & buffer values on both nodes same.

88416 - Zero administration memory management for the ABAP server

2. You can check memory & buffer usage on transaction action ST02. check the over all swaps

BR

SS

[divyanshu_srivastava3]

Hi Samid,

There are lots of docos and wikis and stand practise that you will check, however, I would like to know what are the changes you have done recently on your DB and kernel ?

Regards,

[Sriram2009]

Hi Samid

1. On your ECC system when system performance getting slow or degraded, check the following transaction codes to know the system status

SM66 – To check over all global workprocess status, any table keeps on reading or getting struck?

SM21 – Check the any error log

ST02 – Memory / Buffer usage?

ST22 - Any time out or DBIF ... Dumps?

2. Most of the time Z-reports are running long time. Due to this normal transaction users are getting affected. For this you can enable the transaction code ST12 trace check the overall DB & ABAP program usage.

3. On your Cluster environment one node can have DB & Application instance and another node can handle the end user transactions. Since the windows 2003 server end of support on this year try to upgrade the existing system to either windows 2008 or 2012 with Failover cluster.

4. On your MS SQL DB have you define the 30% of memory for DB? And also you can refer the SCN link how to tune the MS SQL DB

BR

SS

[Former Member]

This message was moderated.