cancel
Showing results for 
Search instead for 
Did you mean: 

horrible response times

Former Member
0 Kudos

Our ECC Ides system today had . For first time 17 users were working on the system (WIN2008/SQL2005 based). Before the people were maximally 5.

The server is done by making a homogeneous system copy from an blade machine(now it is an VIRTUAL)

There was enaught disk space. As I could not log in I had to restart system 3 times to get logged on the system and even then it took almost fewminute to log on.

However I checked Wokload st03n transaction) in system. and found out that

there top abap."Login_Pw", "SESSION_MANAGER", "?". (BAtch), "ADMSBUF, >DEleyed Function call, RSPOWPOO""RSWWclear", ""VA01", "SAPMHHTP "Buf Sync" >DDLOC CLEANUP)""rsbtctE"

What can I do?

¸

What I am most suspicios are 2 jobs. Namely I found 2 jobs canceled running in that period of time SAP_CCMS_MONI_BATCH_DP for 3000+ sekunds (cancelled two times as the system was twice restarted) ,the 2 main users under "users profile" in st03n were ZUGTIN running and SAPSYS( running many system jobs like ."Login_Pw", "SESSION_MANAGER", "?". (BAtch), "ADMSBUF, >DEleyed Function call, RSPOWPOO""RSWWclear", ""VA01", "SAPMHHTP "Buf Sync" >DDLOC CLEANUP)""rsbtctE)?

What else I could look for?

How to analyse/solve the problem for future

What is the role of job SAP_CCMS_MONI_BATCH_DP and how to disable it (I can see it is event triggered and runs every 3 days.

Accepted Solutions (0)

Answers (3)

Answers (3)

Former Member
0 Kudos

We added some RAM

Former Member
0 Kudos

Thank you for your answer. I have 4GB on this virtual machine. Yes, maybe the problem was solved by adding 3 more dialog Wp(to 6 egzisting).

Also may be the issue might be that there was no time for generating load (transaction SGEN) so users encountered this situation(now we are compiling).

However still it seems strange to me that in time frame of bad response time the yob was running having quite large nr of database reads(slow on virtual machine)

Former Member
0 Kudos

so you are running SGEN....that is great

Also I would advice you to disable the job which is causinf the issue,let your server get normal and then again reschedule it....

Let me know of any questions

Rohit

markus_doehr2
Active Contributor
0 Kudos

There are a few points coming to my mind:

- How much memory did you give to your database? If you use the default configuration the database will use 50 % of the available memory which is 2 GB then.

- How big are you buffers (ST02)? Did you change them?

If you did not yet run SGEN, the system will load, compile and save each program that has not yet been executed. This will trash the program buffer (abap/buffersize) and fragment it heavily. If you check ST02 you'll see a lot of swaps because programs need to be re-read again and again from the database.

Virtualized environments seriously suffer from syscalls. A syscall is made everytime the system is doing I/O or transferring data over the network. If you machine is using swap/paging, this will seriously impact performance on virtualized environments.

The increasing number of WPs is not affecting seriously, they use the same (shared) memory pool and just attach to it.

To have a good performance you need to make absolutely sure your database buffers and the SAP buffers fit in physical memory. You can check in ST06 - Detail analysis menu - previous 24 hours - memory if paging of memory is going on. If that is the case, you should tune the memory usage (fixed database memory and check buffers using sappfpar).

You may also set the parameter

es/use_mprotect = false

to decrease the number of syscalls.

Also install the hotfix http://support.microsoft.com/kb/931308.

See Note 1056052 - Windows: VMware ESX Server 3 configuration guideline

Markus

Former Member
0 Kudos

SAP_CCMS_MONI_BATCH_DP is a standard SAP job.It is not a reorganization job but is needed to start tools / methods in the system monitoring area in the background -> transaction RZ20/RZ21.

I dont think you should disable this

Now coming to response times,I guess this is a memory issue

what is the RAM available in your system?

Also you can look to tune your system,look for buffers in ST02 and check if there are too many red entries,if yes you will need to tune this

Also please let us know how many no of work processes are avaialabe in the systen of each type i.e dialog,background and others

You will also need to tune the no of work processes

But main thing is how much memory is there

Let me know alll the information

Rohit