GP CP's SP CP's (was: IBM countersues Neon over zPrime accelerator)

[Patrick Falcone]

Check the link below, there is information on switching in question/answer #2.
 
http://www.itindepth.com/JoseCastano-zAAP.htm
 
A zAAP in my last situation would have beat the freakin pants off a GP CP, 
that's a fact. I'm sitting here with many disparate machines and again find 
myself with another client deliverable to move WAS V7 into situation where a GP 
CP will lose to a zAAP with regards to performance, that's a fact. Do I care 
about overhead switching at that point or relief to support the new workload 
with minimal compromise to the traditional workload? I'll take the zAAP over 
the slower GP CP. 
 

--- On Wed, 2/3/10, Mark Zelden mark.zel...@zurichna.com wrote:


From: Mark Zelden mark.zel...@zurichna.com
Subject: Re: IBM countersues Neon over zPrime accelerator
To: IBM-MAIN@bama.ua.edu
Date: Wednesday, February 3, 2010, 10:37 PM


On Wed, 3 Feb 2010 13:46:24 -0800, Patrick Falcone
patrick.falco...@verizon.net wrote:

Come on Mark, you made a blanket statement about zAAP's and performance and
I disagreed. Now you're coming back with specifics which I'm not going
to/can't argue. I'm just stating I've been in situations where I personally
believe that a zAAP would have been benificial over a GP CP from a
performance perspective.
 
So can we call this a *it depends on the environment* at this point. 
 

I'm not trying to start an argument, but performance is not about beliefs
or gut feelings.  It's about measurements and data / facts to back it up.  

The fact is, that there is overhead in engine switching to move work
over to a specialty processor, so I'm not inclined to believe that
from a performance perspective, overall,  a system would run better with a split
between zAAPs and GPs as opposed to all GPs.   Now, I can see that 
WAS could run better since the zAAP(s) could be sitting there servicing
the java work without competition - but the rest of the system could be CPU
starved (also, remember the additional overhead if these engines are 
in another book).

So I guess my point is, putting aside software costs, it is better from an
overall system performance perspective to have all GPs.  

Mark
--
Mark Zelden
Sr. Software and Systems Architect - z/OS Team Lead
Zurich North America / Farmers Insurance Group - ZFUS G-ITO
mailto:mark.zel...@zurichna.com
z/OS Systems Programming expert at http://expertanswercenter.techtarget.com/
Mark's MVS Utilities: http://home.flash.net/~mzelden/mvsutil.html


--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html

Re: GP CP's SP CP's (was: IBM countersues Neon over zPrime accelerator)

[zMan]

On Wed, Feb 3, 2010 at 7:17 PM, Patrick Falcone 
patrick.falco...@verizon.net wrote:

 A zAAP in my last situation would have beat the freakin pants off a GP CP,
 that's a fact. I'm sitting here with many disparate machines and again find
 myself with another client deliverable to move WAS V7 into situation where a
 GP CP will lose to a zAAP with regards to performance, that's a fact. Do I
 care about overhead switching at that point or relief to support the new
 workload with minimal compromise to the traditional workload? I'll take the
 zAAP over the slower GP CP.


A full-speed CP is the same speed as a zAAP, so your post makes no sense. If
you're still talking about kneecapped CPs, then it still makes no sense. I
can claim that a kneecapped CP on my z10 beats the pants off the IFL from
your MP3000; what does that prove? You're comparing apples and hamsters.

Now, if you had a SINGLE application that could saturate a zAAP, and an
overloaded CP, adding a second CP might not benefit that single application
as much as a zAAP. But that's just adding dedicated resource, not proving
superiority of one over the other. That's what tuning exercises are for.

Apologies if I've misinterpreted your post, but I don't see any other way to
take it.

--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html