HP3000-L Archives

November 1995, Week 2

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
George Reimonn <[log in to unmask]>
Reply To:
George Reimonn <[log in to unmask]>
Date:
Tue, 7 Nov 1995 09:01:27 -0800
Content-Type:
text/plain
Parts/Attachments:
text/plain (37 lines)
>From: mplath
>
>We are a new mpe/ix 4.0 shop.
>The other day we had a batch job use excessive amounts of
>cpu.  Another batch job running at the same time was
>starved for cpu and ran much longer than usual.
>When the first job was suspended the second one was able to
>finish.
>
>My question is:  what, if anything, can be done to
>prevent thissort of situation.
>
>Thanks.
>
>
 
Buy KLA Express from Unisom.  It handles this kind of
thing very well.  When configured properly (which will
take you a while), it will let the two jobs take
turns at stealing the CPU.  I believe you can't run
a shop where users run their own jobs (or different
IS groups that won't coordinate) unless you have this
or something like it.
 
When two jobs are going at it, the job that does more
disk I/Os tends to lose out.  Short of buying KLA, use
the DS and ES queues to try to balance these out.  Put
the I/O intensive job in the DS queue (usually the
default) and put the CPU intensive job in the ES queue
by putting ;PRI=ES on the end of the job card.
 
I use HP's Glance XL to monitor what kinds of resources
jobs are using.  Another must-have for bigger sites.
 
--George Reimonn ([log in to unmask])
IS Director, Providence Medford Medical Center

ATOM RSS1 RSS2