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

Combining two jobs into one will reduce CPU time?


IBM Mainframe Forums -> TSO/ISPF
Post new topic   Reply to topic
View previous topic :: View next topic  
Author Message
bheemas

New User


Joined: 19 Sep 2005
Posts: 6

PostPosted: Mon Oct 17, 2005 4:59 pm
Reply with quote

Hi,

I am bringing again this topic on to the board as some folks are positive in saving CPU time by consolidating the jobs in JCL. Please share your thoughts and advice me. Question goes like this..

In production, we have a scenario where one job solely for doing sort and the other job takes sortout as input and process it. These two jobs are running under two different jobnames (or in Procs).

If we combine these two jobs into one..will there be any savings in CPU time. I was under the assumption, by consolidating like this we can surely reduce CPU as the time taken for i-o operations will come down to some extent. I have tested these jobs separately and consolidated them again and validated. I got mixed results. there by I was not able to come to a conclusion.
I guess we may need to consider Elapsed Time, Active Time, Channel Time also into account when we are refering to CPU time. Please advice me.
_________________
Bheema Sekhar,
Software Engineer,
India
Back to top
View user's profile Send private message
Rupesh.Kothari

Member of the Month


Joined: 27 Apr 2005
Posts: 463

PostPosted: Mon Oct 17, 2005 6:22 pm
Reply with quote

HI Bheema,

Offcource if you consolidate 2 jobs it will take less time if they are inter-related.

I did not get why you get mixed result. It may be some some other reason.

Regards

Rupesh
Back to top
View user's profile Send private message
bheemas

New User


Joined: 19 Sep 2005
Posts: 6

PostPosted: Tue Oct 18, 2005 2:45 pm
Reply with quote

Hi Rupesh,
Thanks a lot for answering. Yes they are related. Sorted output will be given as input to next job. This is the only input the second job takes

But previsouly some one from this forum told we can't save any CPU time but only we can save time in manual submission.

I need to get expert solution asap. I am not able justfy with my test results as i have told you i am getting mixed results.

Regards,
bheemas
Back to top
View user's profile Send private message
Rupesh.Kothari

Member of the Month


Joined: 27 Apr 2005
Posts: 463

PostPosted: Tue Oct 18, 2005 3:06 pm
Reply with quote

HI Bheemas,

As you mentioned that two jobs are related then it will definatly save some time.

Regards

Rupesh
Back to top
View user's profile Send private message
yogesh

New User


Joined: 27 May 2005
Posts: 21

PostPosted: Tue Oct 18, 2005 4:48 pm
Reply with quote

Hi friends,

there is no way u can save the 'CPU' time by combining to jobs in one jcl
even if u r using cobol program with files thru jcl each program will open open and close the file after execution, u can save ur one time when u combine two jobs in jcl so that u need not have to submit thenm twice.
Let me know if i am wrong.
Back to top
View user's profile Send private message
bheemas

New User


Joined: 19 Sep 2005
Posts: 6

PostPosted: Tue Oct 18, 2005 6:29 pm
Reply with quote

Hello,

Here also I am getting mixed results. Who can give us the right suggestion. More over we are submitting jobs using CONTROL-M. In each job hundereds of system variables are set to control the flow. If we have both things in one job we dont need to set these parameters.

Looks like this definately need some brain-storming session and expert speak. Please share your thoughts. waiting for your solution.

Regards,
bheemas
Back to top
View user's profile Send private message
MGIndaco

Active User


Joined: 10 Mar 2005
Posts: 432
Location: Milan, Italy

PostPosted: Wed Oct 19, 2005 1:36 pm
Reply with quote

For what I know, and is not enough, CPU time depend on different variable and so it is not possible run a job twice with the same, exact result.
CPU time depend on your processor, the number of I/O, the amount of stress of machine and many other reason(try to think to a machine that has more than one subsystem).
Remember that what you see as SRT TCP ELAPSED and so on are really far from the CPU time concept(to look the really CPU time you must read the SMF record 40/41.. if I'm not wrong).

I suppose that your scenario need some clarification because if you have a job that produce a file(as sortout or other) and other one that read that file, you have to specify if you read/write from/to tape, dasd or virtual tape, and to have a good test you must run the two solution in the same subsystem first, both from your userid or from production environment, and if is possible with the same priority.
After this you will have the two time that have few differences(but they always have!).

If this can't help you I suppose that you must put some feedback of your execution to help us.

I hope in this.
Back to top
View user's profile Send private message
phil rook

New User


Joined: 25 Nov 2005
Posts: 12
Location: amsterdam

PostPosted: Fri Nov 25, 2005 10:01 pm
Reply with quote

Adding steps together will not make any real difference.
Steps should be added together when the steps are related, i.e. output from one step is input to other.
You can reduce sort time sometimes by adding more SORTWK files.
Dynamic allocation is usually marginally slower.
Back to top
View user's profile Send private message
k_vikram07

New User


Joined: 23 Nov 2005
Posts: 35

PostPosted: Sat Nov 26, 2005 11:04 pm
Reply with quote

I am in a thinking that combining two jobs into one will make the execution of the JOB faster .

My reasoning was, the allocation of the datasets by the JES. JES may have to allocate the datasets twice, if it were in two jobs.

Let me know if this doesn't affect.

I also understand that the CPU time for each of these STEPS may not be any different.

Thanks
Vik.
Back to top
View user's profile Send private message
phil rook

New User


Joined: 25 Nov 2005
Posts: 12
Location: amsterdam

PostPosted: Sun Nov 27, 2005 9:55 pm
Reply with quote

to be quite honest I do not know why you are wasting time on this - you will not make any significant saving.

More important things are using the correct block size for the dataset ( this may reduce the CPU time),

or if the output from the sort is on tape using vol= (,retain).

Experiment using different block sizes - on some systems five blocks to a track works better. This was to do with Ibm machines having 5 buffers available for I/O. Whether on the newer Cpu's this is different I do not know.
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 -> TSO/ISPF

 


Similar Topics
Topic Forum Replies
No new posts Finding and researching jobs All Other Mainframe Topics 0
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 How to create a list of SAR jobs with... CA Products 3
No new posts Help in Automating Batch JCL jobs mon... JCL & VSAM 3
Search our Forums:

Back to Top