RE: Interactive Feature Upgrade Theory

Chris,

Let me rephrase your question if I may:

What are you willing to do for $160,000 USD?

This dollar figure represents the difference in list price on an 830
4-way between 240 and 560 Interactive CPW (price should be the same
regardless of model).  Even assuming you would get a discount of some
sort, it is still a sizable figure.

For that amount of money, would your management be willing to invest a
smaller figure in training/analysis/coding to get the interactive CPW
figure to a lower amount?  That's my first reaction when you mention
that some of your interactive load is query users.  It's really a
management kind of decision, assuming that they can get appropriate
information.  Workload growth on the machine would also be factor, do
you expect an increased application load in the near future?

If the applications are the main part of the problem and they don't lend
themselves to easy analysis, I would try to budget for the additional
interactive CPW as an upgrade after initial installation.  If you move
from 'constantly' reaching the threshold to 'occasionally' reaching the
threshold, is that okay?  If not, then have the budgeted upgrade
installed.

An additional consideration is the way the newer hardware/OS behaves
when the threshold is reached.  Check the appropriate versions of the
Performance Capabilities Reference Guides for specifics.  My
understanding is that with each generation, they have lessened the
impact of the CFINT jobs on the batch CPW.  In other words, your
interactive will slow down at the threshold, but your batch will not be
impacted as dramatically.  That's the publicity anyway.

IBM Performance Guides:
http://www-1.ibm.com/servers/eserver/iseries/perfmgmt/resource.htm

I think that your logic is reasonable, assuming that CPU speed is the
limiting factor on your intense interactive applications.  If each job
executes in a shorter amount of time, and the number of jobs doesn't
change, then the probability of those jobs running concurrently
decreases.

I don't think that the effect of more powerful processors is linear to
job duration however.  IBM wrings additional CPW out of their newer
processors in a variety of ways.  Whether this additional power
translates significantly for your system depends on what you're doing.

So I would expect some decent improvement in interactive run times, but
your application mix will be the prime determinant as to how dramatic
the improvement is.

Regards,
Andy Nolen-Parkhouse

> On Behalf Of Chris.Whisonant@comporium.com
> Subject: Interactive Feature Upgrade Theory
>
>
> We are a pretty heavy interactive company currently on a 720 2-way
V5R1
> with 240 CPW Interactive. We constantly reach that threshold and in
comes
> the CFINT troopers (2 CFINT jobs taking up nearly 40% total CPW at
> points).
> Part of this is due to programming (apps developed by a subsidiary)
and/or
> training. We've gotten many query users to submit to batch, but some
still
> like to just get the results on screen.
>
> We are going to upgrade around year-end (820 or 830 4-2ay) and I've
got
> some questions. Obviously the Interactive Feature is only giving you a
> certain amount of CPW. On our current system, some interactive jobs
can
> use
> up to (or over) 10% CPW which equals around 80 CPW. If we upgrade to a
> much
> faster system one would assume that these jobs may only use the same
> amount
> of cpw and not the percentage. But these jobs should also take far
less
> time to run. The dilemma is whether to stick with our 240 CPW or pay
our
> fair share of the "tax" and step up to 560 CPW.
>
> Any advice or words from experience are appreciated,
> Chris Whisonant