RE: QZDASOINIT question

Joe:

You're right.  CFINT is only triggered for jobs that are performing green
screen IO - local or remote attach, telnet, or pass-through - only for 5250.
The trigger is CPU utilization.  The trigger is NOT transactions per unit
time, number of logged-on users, or number of active or attached or
configured devices.

The ODBC jobs play no part in triggering CFINT.  There will always be some
CFINT but I can run a large system to 95 percent CPU utilization with
QZDASOINIT and never have any significant showing from CFINT.

For the purpose of this explanation, assume that you need to support 100
users.  Determine the required interactive capacity using a procedure
something like this.

1. Get all of your users ready to work but waiting for a phone call from
you.
2. Have 5 users start to work.
3. Watch CFINT using WRKSYSACT.
4. If CFINT is less than 10 percent, add more users until CFINT is
consistently hitting 10 percent.
5. That number of users is about the maximum number that your system will
support.  For the sake of argument, assume that CFINT hits 10 percent with
12 users.
6. Divide the total number of users by the discovered number.  100 divided
by 12 equals 8.3.
7. Multiple your current interactive rating by 8.3.  That is a simple
prediction of the required interactive CPW.

Be careful with workload mix, month-end peaking, and the law of small
numbers.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Richard Jackson
mailto:richardjackson@richardjackson.net
http://www.richardjacksonltd.com
Voice: 1 303 808 8058
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

> -----Original Message-----
> From: owner-midrange-l@midrange.com
> [mailto:owner-midrange-l@midrange.com]On Behalf Of Joe Teff
> Sent: Friday, August 04, 2000 9:04 PM
> To: 'MIDRANGE-L@midrange.com'
> Subject: RE: QZDASOINIT question
>
>
> There is an AS/400 that runs both green screen apps and supports
> client/server.
> There are system slowdowns that are being questioned. I have seen periods
> where the CFINT01 task is using 60%-80% of the CPU. The vendor of
> the green
> screen app is claiming that the ODBC jobs are the cause. I say
> that the ODBC
> jobs may by adding to the problem, but are not themselves
> responsible in any
> way for the governor's activity. I think they need to upgrade
> their system so
> they have more interactive CPW.
>
> On Friday, August 04, 2000 8:25 PM, Pete Hall
> [SMTP:pbhall@execpc.com] wrote:
> > At 23:21 08/02/2000 , Joe Teff wrote:
> > >Is there any way that ODBC requests coming in through some QZDASOINIT
> > >pre-start jobs could be considered as interactive by the
> governor? I always
> > >thought that these jobs were the essence of server jobs.
> >
> > I guess I don't really know the answer to that, but what's the
> reason for
> > your question? The last place I worked was a heavy user of ODBC, and we
> > modified the configuration so that ODBC ran in its own pool at
> priority 35
> > and with a shortened timeslice, because as shipped, it was
> killing response
> > time. Once that was done, overall throughput improved
> considerably. This
> > was on an S30 and later a 730.
> +---
> | This is the Midrange System Mailing List!
> | To submit a new message, send your mail to MIDRANGE-L@midrange.com.
> | To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
> | To unsubscribe from this list send email to
> MIDRANGE-L-UNSUB@midrange.com.
> | Questions should be directed to the list owner/operator:
> david@midrange.com
> +---

+---
| This is the Midrange System Mailing List!
| To submit a new message, send your mail to MIDRANGE-L@midrange.com.
| To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
| To unsubscribe from this list send email to MIDRANGE-L-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator: david@midrange.com
+---