How to moniter the max quantity of used dialog work process in SAP system

[Former Member]

Dear Gurus,

The quantity of Dia Work Process(DWP) of APP1 is: 16 (rdisp/wp_no_dia=16)

The quantity of Dia Work Process(DWP) of APP2 is: 16 (rdisp/wp_no_dia=16)

The quantity of Dia Work Process(DWP) of APP3 is: 20 (rdisp/wp_no_dia=20)

The quantity of Dia Work Process(DWP) of APP4 is: 20 (rdisp/wp_no_dia=20)

So the quantity of DWP in our SAP system is: 72 (=161620+20).

We have 75 dialog users and many RFC callings via JCO method from J2EE system.

I'm afriad the 72 DWPs will be run out of use someday.

So I want to moniter the max quantity of used DWP every per hour to prevent the situation I mentioned above.

Could you give me some suggestions?

Thank you very much..

Jason

[Former Member · Accepted solution]

Hi,

You can also use SM50 on each app server and click on the "CPU" button (CtrlMajF6).

If you have enough dialog work processes, some of them should be nealy never used and so should have a CPU time of a few seconds only.

My production ECC6 system has the CI + 6 DIA servers for a total of 235 dialog work processes.

I checked SM50 on one of the app server : after the sunday weekly restart, the most CPU intensive DIA process has consumed 2h42mn10 s of cpu time. The last five DIA processes have consumed less than 5 seconds of cpu time.

Therefore I know that I have enough DIA processes in my system.

Regards,

Olivier

[Former Member · Accepted solution]

Hi Olivier,

I really appreciate your help. You told me a very useful clew!

As the method you told me, I have some questions:

(a3) Do you restart your 'CI + 6 DIA' separately every week? Why did you do that?

(a4) How to calculate 'CPU intensive time'? For example, App server 'APS' was restarted on 8-10-2011 and stopped on 8-20-2011.

There was a DWP(PID 557550) was created and activated on APS during this period(10 days). So If I add up each 'cpu time' of the task

which 557550 executed, I can get the 'CPU intensive time' of 557550 (dispalyed in SM50). Right?

In other words:

DWP(PID 557550) finished task1: cpu time = 5ms 8-10-2011

DWP(PID 557550) finished task2: cpu time = 4ms

DWP(PID 557550) finished task3: cpu time = 6ms 8-20-2011

So the 'CPU intensive time'(Between 8-10-2011 and 8-20-2011) of DWP(PID 557550) is: 15ms (5ms4ms6ms). Right?

(a5)Can we moniter the 'CPU intensive time' of DWPs in CCMS automatically? (I want to prevent the situation: The DWPs in our entire ECC are occupied.)

Thank you!

Jason

[Former Member · Accepted solution]

Hi Jason,

(a3) we restart our our CI and 6 DIA each week because we take an Oracle offline backup each week.

(a4) I don't understand what you mean. With SM50 you can just display the amount of cpu time consumed by each process since the last restart. If you want to know the cpu time consumed for a specific task, you should better use transaction STAD.

(a5) Sorry, I don't know. I don't use CCMS. I don't believe in automatic moniitoring.

Regards,

Olivier

[ashish_vikas]

I think you are afraid for a situation that all 72 DWP are occupied and you have no free workprocess to process incoming request.. if yes, you can get this info from ST03N. If in any case, all wps gets occupied, your avg. wait time will be very high for that particular hour.

ST03N > expert mode > workload > timeprofile > task type > DIALOG.

generally avg. wait time is only few ms. but for this problem, it will be high.

thanks

ashish

[Former Member]

Hi Ashish

Thank you very much! You let me know a very usefull moniter way.

As to 'avg. wait time', I have two questions:

(a1) Supposing 'avg. wait time'< x OR 'avg. wait time'< y% * 'avg resp. time/Dialog steps' could measure the entire using status of DWPs is healthy. Do you know the specific value of 'x' or 'y' ?

(a2)Do you know how to moniter 'avg. wait time' in CCMS?

Thank you!

Jason

[ashish_vikas]

(a1) Supposing 'avg. wait time'< x OR 'avg. wait time'< y% * 'avg resp. time/Dialog steps' could measure the entire using status of DWPs is healthy. Do you know the specific value of 'x' or 'y' ?

> it will ve very less, only few milisecs for normal operations (in my system 10ms/DIA step). you can check your system for a hour when there was no problem. (you have the option to verify it for application server wise)

http://help.sap.com/saphelp_nw70ehp1/helpdata/en/c4/3a6af1505211d189550000e829fbbd/frameset.htm

> (a2)Do you know how to moniter 'avg. wait time' in CCMS?

No, sorry.

thanks

ashish