A checkpoint counter sets at 100,needs to reduce time of job

N.Kotivamsi · New User Joined: 25 Jul 2012 Posts: 4 Location: India

Dear All,

I have a program which reads the files and creates one checkpoint-id after completing at evry 100 records(Checkpoint counter which i've given in checkpoint file),I need to do one checkpoint-id in one second.
Could you please advise on this.Many thanks in advance !!!

Please let me know if you need any other details.

Thanks !!!

dick scherrer · Posted: Wed Jul 25, 2012 9:00 pm

Hello and welcome to the forum,

Why does someone believe a checkpoint is needed every second?

It is possible that NO checkpoints are needed at all. . .

Anuj Dhawan · Posted: Thu Jul 26, 2012 12:55 pm

First, this question is very site-specific. Every shop will have its own style of writing such procedures -- suggest you ask some one around how it is done at your shop.

Second, most of the shops, I had been to, using check-point used to take a 'snapshot' of the current application state, and later on, use it for restarting the execution in case of failure. And time was never a parameter of snap-shots. So as Dick says, why a checkpoint is needed every second?

Anuj Dhawan · Posted: Thu Jul 26, 2012 12:56 pm

Just in case, You might like to check this link: Ranked Search Results for Document: dfsapgg1 "IMS V9 Appl. Programming: Design Guide"

N.Kotivamsi · New User Joined: 25 Jul 2012 Posts: 4 Location: India

Hi Scherrer/All,

Thanks for your warm welcoming wishes and quick reply on this!!!

My Program is Batch IMSDB program and the checkpoint that we are using in program is Extended checkpoint,Please find the request that i got from DBA team,

Issue: Please change the checkpoint count for program XXXXXXXX

It is currently doing only 290 checkpoints on 11 minutes of wall time. It should Be 1 checkpoint a second. This will hopefully reduce the DFS2389I [MQ to IMS messaging] Buffer Full Condition.

please see the below sample of time difference b/w 1st and 290th chkpt-id,

PROGRAM=XXXXXXXX,CKPTID=XX000001,REGID=xxxx,DAY=199,TIME=0532045 --------------- 1st chkpt-id
DAY=199,TIME=053204596302-0400 X2P2

PROGRAM=xxxxxxxx,CKPTID=xx000290,REGID=xxxx,DAY=199,TIME=0543384 ----------------290th chkpt-id
DAY=199,TIME=054338410302-0400 X2P2

Kindly advise me on how to start on this request and let me know if any other details needed.

Thanks !

Anuj Dhawan · Posted: Thu Jul 26, 2012 2:55 pm

N.Kotivamsi · New User Joined: 25 Jul 2012 Posts: 4 Location: India

Hi Anuj,

I've shown few of my peers at my place,They have advised me to reduce the checkpoint counter from 100 to 35.Not sure whether it will work or not and not sure what will be the impact after change.So i thought it's good to get some ideas from fourm...

Thanks!

PeterHolland · Posted: Thu Jul 26, 2012 5:40 pm

dick scherrer · Posted: Thu Jul 26, 2012 6:40 pm

Hello,

Sounds like "you" (collectively - apps, database, mq, whoever) need someone with a lot of experience with MQ/IMS applications to review the design and see if there is a flaw with the way this is implemented.

Increasing the frequency of the checkpoints may or may not help the "problem", but it surely might cause others. . .

Anuj Dhawan · Posted: Thu Jul 26, 2012 6:49 pm

N.Kotivamsi · New User Joined: 25 Jul 2012 Posts: 4 Location: India

Hi All,

Thanks for all giving quick replies on this !!!

@Peter,

Could you please eloborate your comments.Thanks in advance !!!

Thanks!

PeterHolland · Posted: Fri Jul 27, 2012 1:57 pm

What can i say? The manuals say it all about that message. A DBA may know a lot about databases, but doesnt need to know about the inners of IMS.

don.leahy · Posted: Fri Jul 27, 2012 5:47 pm

IMO one checkpoint per second is not that unreasonable if the goal is to minimize contention and/or buffer usage. I would hope that in making this recommendation that your DBAs have considered the impact this increased frequency will have on the IMS logs. But I agree with Peter that the IMS sysprogs need to be in the loop as well.

Ed Goodman · Active Member Joined: 08 Jun 2011 Posts: 556 Location: USA

So the buffer can hold a few seconds worth of messages??

It could be the messages aren't released until the commit point (depends on set up), so increasing the commit frequency will make for smaller bunches.

It's weird that it was presented as a time span issue though. It really does sounds like a volume issue. It probably should have been presented as "the queue fills up after 5,000 records, so be sure to release them before that."