on 06-13-2009 8:46 AM
Hello,
We have our production system as below:
OS : Windows 2003 Sp2
Database : IBM DB2 v 8
SAP : ECC 6.0
RAM : 12 GB
CPU : 8
The performance of the system is very good when we stop and start the SAP instance (improves more with the OS restart).
So, after a restart I believe that the memory/swaps is cleared. As time goes by the swap defragmentation takes place which maybe causing this slow performance.
Can any one suggest how to diagnose this problem and what could be the possible resolution.
Thanks very much !
Resolved through work around
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
The performance of the system is very good when we stop and start the SAP instance (improves more with the OS restart).
As well as stopping the instance have you stopped the SAP related services in the Windows System manager, as these can also hold memory, if you try this and notice that it is as good as a OS restart then you will know it is in the setup of SAP.
Do you see any Swaps in ST02 because I would expect that this is the issue or a dispatcher is not committing which we have seen and hence not releasing memory, this should be able to be seen in Windows under task manager then just line the PID up in sm50.
I wouldn't have thought 8000 is high, I'm just applying the swapping page hotfix and we are current swapping at 54K.
How many dispatchers are you running?
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hello All,
Thank you for your replies and sorry for the late reply.
The issue has been traced to the OS level (not SAP).
We are running Windows 2003 enterprise SP2, and the Filecache is consuming a lot of memory which it does not free free up, so eventually at the end of 2 - 3 days we are left with no free physical RAM. For now we have devised a work around i.e. to use a small utility called cacheset from sysinternals (microsoft site) that frees up the file cache so that the physical memory is freed.
Thanks again for all those who gave their vvaluable inputs..
Cheers!
for windows , you should take care page in , not page out , and could you show what's parameters for your memory setting ? did you use the ZERO memory management ?
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hello,
All the parameters have been changed according to ZAMM. We increased the physical RAM. What happens is the available RAM at the OS level goes on decreasing and is never freed up, so when the system is restarted it frees up and the users can work.
How can we diagnose and correct this problem.
Thank you
i have had this problem but with SQL server, after time disp+work didn't free the used memory and i got memory issues. What i did was to kill restart the WP, so the memory gets freed. (of course you do this only with free WP)
Not the best way, but still works, without having to restart the entire instance
Hello,
Thanks for the replies..
Killing and restarting the work process is not really a solution ( but is handy during pressure situation ).
What I feel is this problem of memory consumption can be caused either by the OS or by the application considering that there are no other applications running on that server.
So how can we first find out whether this issue is with the OS or SAP application.
Thanks
Zaheer,
Check SAP Note 1009297 -Windows Server 2003 Family: High Paging Rates
Thanks
Sushil
Edited by: Sushil Suryawanshi on Jun 18, 2009 4:21 PM
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Zaheer,
You need to moniter your system & do some fine tune few parameters.
Goto ST06 & check the CPU Utilization during peak hours.Compare daily utilization & Check for Paging.
Also Monitor ST02, ST03 & ST07 tcodes.
ST02->Check Buffer Swaps & memory usages
ST03-> Check Avg Response time
ST07-> Check Response time
Hope it helps
Thanks
Sushil
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hello,
Thank you for the response.
I have studied the standard monitoring and performance transactions. The st02 shows hit ratios of 99%.
ST04 shows a buffer quality of 95 %.
The only doubtful thing I saw was that the pages out/second is ver high in st06 (between 3000 to 8000)
But as mentioned earlier, once the instance is restarted the performance is very good and in 20hrs its very slow.
Pplease suggest.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.