I am actually typing this for the second time.  The first time I typed it,
by the time I had hit ‘SEND’, my user login had timed out and my page with
all of my typing had “expired”.  So here I go again (this time I’m typing
in Word and then pasting):

I somewhat agree with ‘dbrinson’ below.  My setup just goes one important
step further.  The setup I am about to describe looks strange, but if you
read the whole text, you will come to see that it makes good sense.  I have
used this setup for 14 years and it has always worked well.

CQ – 155, 255,decay
DQ – 190,240,decay
EQ – 150,154,decay

The CQ is intended for online users.  But sometimes, programmers
or “runaway” processes can use large amounts of CPU.  A “normal” online
user will always remain below 190.  Most online processes run between 155
and 175.  Therefore, “normal” online users are not affected by batch
processes, which are located above 190.  If however, a process uses a lot
of CPU, it slowly falls to the 255 level.  This way, it has been relegated
to the old EQ.  It ends up having the lowest priority, behind online users
and batch.  You must adjust your time-slice and quantum parameters on the
CQ depending on the speed of your system.  I have always set them so that
it takes a CQ process 30 CPU seconds to get from 155 to 255.  I have always
thought that no online process should need more than 30 CPU seconds to
complete.  If it does, it must be some kind of CPU intensive report and it
should be done in batch.

The DQ is for batch processes and is set to have priority behind “normal”
online users, but ahead of “bad” online users.  This way, batch jobs do not
interfere with “good” online users and they are not affected by “bad”
online users.

The EQ is for systems people, such and System Manager, Operator, DBA, etc.
They have top priority over everybody.  Should something go wrong, you want
to be sure your systems people can get a “:” prompt and are able to get CPU
time to fix things.

In summary, systems people have top priority, followed by “normal” online
users, followed by batch, and then finally online “abusers”.

I have been using this setup for well over 10 years now and it has always
worked well, if I do say so.  I first developed it when I was working on a
Series 950 with 100+ online users, batch reports, and 5 programmers all
banging on the system at the same time.  Needless to say, the users were
unhappy because the programmers constantly sucked up the resources.  And
when I told the programmers to put compiles and Query searches into batch,
they ignored me.  And even when they did run in batch, they clogged up the
job queue.  So, I came up with this setup.  The users were happy.  The
batch jobs were happy.  The programmers were the only ones complaining now,
but not that much.

All I can say is try it and see if it works for you.  It automatically puts
each process in the correct place in the queue and keeps it there.

Randy Keefer

These are my opinions and not my employers or anybody else’s.