IBM Mainframe Forum Index
 
Log In
 
IBM Mainframe Forum Index Mainframe: Search IBM Mainframe Forum: FAQ Register
 

Job running time problem


IBM Mainframe Forums -> ABENDS & Debugging
Post new topic   Reply to topic
View previous topic :: View next topic  
Author Message
luban

New User


Joined: 19 Nov 2005
Posts: 26
Location: Shanghai, China

PostPosted: Mon Aug 04, 2008 9:12 am
Reply with quote

Hi,

I have an question about job running time. Is there anyone could give me any advise? Thanks.

I ran a job twice in one day, but 1st time it just cost 6 mins, 2nd time it cost almost 44 mins.

I've checked JESYSMSG of the job and found the step which cost the time difference.

I found the difference only is STEP SERVICE:

STEP SERVICE 15,043,989
IEF373I STEP/PS010 /START 2008214.0841
IEF374I STEP/PS010 /STOP 2008214.0846 CPU 1MIN 09.23SEC SRB 0MIN 00.48

STEP SERVICE 16,251,120
IEF373I STEP/PS010 /START 2008214.1414
IEF374I STEP/PS010 /STOP 2008214.1458 CPU 1MIN 15.51SEC SRB 0MIN 00.56

So is there anyone could give any advice about " STEP SERVICE " or any other aspect I missed?

Thank you very much. " )
Back to top
View user's profile Send private message
dick scherrer

Moderator Emeritus


Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix

PostPosted: Mon Aug 04, 2008 10:53 am
Reply with quote

Hello,

You have observed a difference in elapsed time. Elapsed time is not typically used to measure the cost of a job.

The difference in elapsed time is most likely due to the difference in what else was running on the system or why the second job had to wait for something (like a tape mount).

If you notice the other statistics presented (step service, cpu, srb) they are quite close. These more correctly point to the "cost" of the job.

It is common to see very different run times caused by other activity in the system.
Back to top
View user's profile Send private message
luban

New User


Joined: 19 Nov 2005
Posts: 26
Location: Shanghai, China

PostPosted: Mon Aug 04, 2008 12:06 pm
Reply with quote

Hi dick scherrer,

Following is the other statistics, actually most of them are really close, except STEP SERVICE.

Do you have any idea about this?
(Actually here's no contention message in the log, so I'm really confused icon_cry.gif )
Back to top
View user's profile Send private message
dick scherrer

Moderator Emeritus


Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix

PostPosted: Mon Aug 04, 2008 12:19 pm
Reply with quote

Hello,

Quote:
Actually here's no contention message in the log, so I'm really confused
No, there typically will not be. . . The system is set to "take turns" between processes that are competing for resources. It does this automatically. What happens is that the more things that are running, the slower some things will run. Part of this is due to volume and part of this is due to system priorities.

From what you've posted, things appear to be quite normal. The runs are very similar - the second job using slightly more resources that the first. For a deeper carification you would need to talk with your performance measurement people.

As i mentioned earlier, elapsed time is not as predectable. Many things can influence run-time/wall-time.
Back to top
View user's profile Send private message
luban

New User


Joined: 19 Nov 2005
Posts: 26
Location: Shanghai, China

PostPosted: Mon Aug 04, 2008 12:32 pm
Reply with quote

Hi d.sch,

Thank you very much for your advice. icon_smile.gif
Actually I will talk with my SME tonight.
It's really weird that just slightly difference can cause 44 mins delay...

Anyway, thank you very much icon_smile.gif
Back to top
View user's profile Send private message
dick scherrer

Moderator Emeritus


Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix

PostPosted: Mon Aug 04, 2008 12:39 pm
Reply with quote

Hello,

Quote:
It's really weird that just slightly difference can cause 44 mins delay...
The slight difference is not causing the difference in run times.

The load on the system at different times is most likely causing the difference in run-time.

If your system has very little going on over the weekend you might try to run the job a few times (with all of the exact same data) and see the run-times and other resource usage.
Back to top
View user's profile Send private message
Anuj Dhawan

Superior Member


Joined: 22 Apr 2006
Posts: 6250
Location: Mumbai, India

PostPosted: Mon Aug 04, 2008 1:00 pm
Reply with quote

Quote:
It's really weird that just slightly difference can cause 44 mins delay...

You might try by requesting Performance Monitoring Team to set up strobes in your JOB..
Back to top
View user's profile Send private message
luban

New User


Joined: 19 Nov 2005
Posts: 26
Location: Shanghai, China

PostPosted: Mon Aug 04, 2008 5:31 pm
Reply with quote

Hi Anuj D,

Strobes?
Could you give me more explanation about this?

Thank you very much.
Back to top
View user's profile Send private message
expat

Global Moderator


Joined: 14 Mar 2007
Posts: 8797
Location: Welsh Wales

PostPosted: Mon Aug 04, 2008 5:39 pm
Reply with quote

Not too much difference in the service units or CPU times used, could well be external influences.

Is this a cloned run using EXACTLY the same input data or with different input data in the second job run ?
Back to top
View user's profile Send private message
luban

New User


Joined: 19 Nov 2005
Posts: 26
Location: Shanghai, China

PostPosted: Mon Aug 04, 2008 5:47 pm
Reply with quote

Hi expat,

Yes, it's EXACTLY the same input data in the second job run.

Thanks.
Back to top
View user's profile Send private message
dick scherrer

Moderator Emeritus


Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix

PostPosted: Mon Aug 04, 2008 7:26 pm
Reply with quote

Hello,

You are spinning your wheels . . .

You cannot "tune" elapsed time. You can (sometimes) improve elapsed time by tuning actual resource consumption when one task uses more resources than another.

Your 2 tasks use basically the same amount resourses. The way to get the same elapsed time for multiple runs is to control the mix of processes that are running on the computer to always be the same when this process runs.

As i mentoned earlier, plan on running the same process multiple times over a weekend or some other time when system use is very low.
Back to top
View user's profile Send private message
Terry Heinze

JCL Moderator


Joined: 14 Jul 2008
Posts: 1249
Location: Richfield, MN, USA

PostPosted: Mon Aug 04, 2008 7:36 pm
Reply with quote

dick scherrer wrote:
...You cannot "tune" elapsed time...
You can stop all initiators except the one you use. icon_lol.gif Just kidding; please don't try this at home. icon_smile.gif
Back to top
View user's profile Send private message
expat

Global Moderator


Joined: 14 Mar 2007
Posts: 8797
Location: Welsh Wales

PostPosted: Mon Aug 04, 2008 8:05 pm
Reply with quote

Quote:
Just kidding; please don't try this at home. icon_surprised.gif

Much more fun trying at work instead .......................
Back to top
View user's profile Send private message
Anuj Dhawan

Superior Member


Joined: 22 Apr 2006
Posts: 6250
Location: Mumbai, India

PostPosted: Tue Aug 05, 2008 5:44 pm
Reply with quote

Hi,
luban wrote:
Strobes?
Could you give me more explanation about this?
From the thread's point of view, I think, "STROBE" should not create this exclamation, I doubt it's a site-specific term because I've used it with IBM personnels while was tuning my JOBs for performance.

Please search on "strobe in mainframes job" in google you will find some useful hits.

And if you might have searched in this forum, you could have find these-

www.ibmmainframes.com/viewtopic.php?t=14734&postdays=0&postorder=asc&start=15
www.ibmmainframes.com/viewtopic.php?t=23354&highlight=stobe
www.ibmmainframes.com/about15214.html

Here are some documents-
ibmmainframes.com/viewtopic.php?t=13272&start=0&postdays=&postorder=&highlight=strobe


You might try TSO STROBE, it invokes "STROBE OPTIONS" scren for me at my site, but I don't use it & just let IBM performance monitoring team to use it..


PS. Any how lost the track of the thread, so a little late reply.. icon_redface.gif
Back to top
View user's profile Send private message
dick scherrer

Moderator Emeritus


Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix

PostPosted: Tue Aug 05, 2008 9:16 pm
Reply with quote

Hello,

Using Strobe to "watch" your job will most likely just waste your time. Strobe could point out bottle-necks in your process, but they would most likely be consistent rather than what you have seen.

One more time. . . The reason you see different elapsed times is most likely due to other activity going on in the machine. Your code used nearly the same amount of resource. If the resource used varied by some large amount, that might account for the difference in elapsed time.

The resource usage from both runs was quiite similar, so the difference in run time is probably not due to the actual process.

One thing to look at might be if this process uses some shared vsam or database data. If other processes heavily hit either, there will be an impact on the process you are interested in.
Back to top
View user's profile Send private message
Anuj Dhawan

Superior Member


Joined: 22 Apr 2006
Posts: 6250
Location: Mumbai, India

PostPosted: Fri Aug 08, 2008 4:50 pm
Reply with quote

Hi Dick,

Perhaps I got the context of the thread little different, per you, please let me know when strobe should be set-up for a JOB?
Back to top
View user's profile Send private message
dick scherrer

Moderator Emeritus


Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix

PostPosted: Fri Aug 08, 2008 8:46 pm
Reply with quote

Hi Anuj,

Strobe is quite good at helping to identify bottlenecks.

If a process consistently preforms poorly, Strobe might be used to isolate where various concentrations of cpu or i/o is being used.

The job for this topic does not seem to fall in this category. With the same input, it runs for different lengths of wall-time. It does not use (from what we've seen) different amounts of resources. Which leads me to believe the difference in elapsed/wall tme is due to external influences rather than something in the code of "this" process.
Back to top
View user's profile Send private message
View previous topic :: :: View next topic  
Post new topic   Reply to topic View Bookmarks
All times are GMT + 6 Hours
Forum Index -> ABENDS & Debugging

 


Similar Topics
Topic Forum Replies
No new posts Running REXX through JOB CLIST & REXX 13
No new posts Map Vols and Problem Dataset All Other Mainframe Topics 2
No new posts To get the the current time DFSORT/ICETOOL 13
No new posts RC query -Time column CA Products 3
No new posts C Compile time time stamps Java & MQSeries 10
Search our Forums:

Back to Top