I usually stress that adding zIIPs and zAAPs provides a performance
benefit for both of the reasons that Bill mentions below. The most
significant for everyone is the second benefit (reducing the strain on
the CPs). But you need to realize that the 'notable exception' is
getting to be the rule. For example, the z10-BC zIIP/zAAP is the
speed of a z01 (about 700 MIPS) for all 130 models, including the A01
(30 MIPS for a UP). If you were running DB2 on an A01, wouldn't you
prefer to run DB2 on a 700 MIPS zIIP than a 30 MIPS CP? I think it's
a no-brainer, especially since the cost of the specialty processors is
lower than the regular CPs.
Be sure to run WSC's zPCR to determine what you can expect to see in
your installation.
Cheryl
On Nov 17, 2008, at 9:52 AM, Bill Seubert wrote:
On Wed, 12 Nov 2008 12:42:31 -0500, Tony Harminc <[EMAIL PROTECTED]>
wrote:
OK, OK - you (and IBM) win! ...
Tony, to the best of my knowledge, no one in IBM System z marketing,
or for
that matter, anyone with significant knowledge of the technical
aspects of
System z, has made official, public statements about specialty
processors
being features to boost performance. There may be a well-meaning
sales rep
or specialist or press person out there who does not have a full
understanding of the hardware who have made that claim, but it should
not
have been an official IBM claim.
As has already been stated, there is one notable exception to the
"Specialty
engines are not performance enhancers" rule - machines that run at
subcapacity. If you have a box that doesn't run at the fully-rated uni
speed, a specialty engine will provide better performance.
There's one other "performance benefit," but it is a roundabout way of
claiming that the specialty engine provides improved performance - if
one
were to install a zIIP or zAAP and relieve the general purpose CP pool
of a
CPU bottleneck, then that would indirectly result in a performance
benefit
by offloading Java and/or other MIPS and relieving the constraint on the
GPs. Thus you got cheaper MIPS with the zAAP/zIIP and "fixed" a
performance
bottleneck. But that's obviously stretching things...
FWIW.
----
Bill Seubert
System z I/T Architect
IBM Corp
[EMAIL PROTECTED]
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
