Subject: | |
From: | |
Reply To: | |
Date: | Tue, 24 Jul 2007 10:04:55 -0700 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
I think the copy of the JMS manual that I have pre-dates the HP "jobq" feature where you can create separate queues and apply limits where needed, so I'm curious: how does JMS work with HP's jobqueues, if at all?
We have a "situation" where I'm currently employed where they are using JMS to "single thread" a particular process. Unfortunately, the "wakeup and check" time for JMS is every 30 seconds. I say "unfortunate" as the jobs in question take less than 10 seconds to complete, and in some cases the users submit them at a rate of 6-12 per minute -- JMS becomes a bottleneck with severe consequences.
At 12 jobs/minute, the "backlog" of jobs would theoretically only grow by a minute per minute, however with JMS in the way the backlog grows at 5 minutes per minute. Once the users stop, the queue would clear at a rate of 6 jobs/minute, but JMS would keep that throttled back to 2 jobs/minute -- in 15-20 minutes, users can create a complete mess :) In effect, JMS is making each job run 3 times longer than usual
So, does anyone know if I can use both JMS and HP jobq's effectively? I'm thinking that I'd still use JMS to qualify the job, and perhaps from that determine/set the "MPE job queue" for the job and then release it accordingly. MPE would then deal with single-threading the actual execution of these jobs with no delay between each job in this "queue"
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|
|
|