Re: Rule of thumb about paging / page fault rate
- From: mark.zelden@xxxxxxxxxxxx (Mark Zelden)
- Date: 22 Dec 2005 10:36:57 -0800
On Thu, 22 Dec 2005 12:14:01 -0500, Knutson, Sam <SKnutson@xxxxxxxxx> wrote:
>Hi,
>
>In a related topic does anyone have thoughts or recent references to
>settings for MCCAFCTH in PARMLIB IEAOPTxx?
>
>I currently use MCCAFCTH=(5000,6000), on my OLTP production partitions
>with 32M and very large AFQ (no paging) and still have
>MCCAFCTH=(2000,2500), for my test and development partitions which range
>from 3M for a sandbox to 22M (with only a little paging) for the largest
>overloaded parking lots for development CICS/DB2/IMS/MQ subsystems. We
>arrived at these values circa OW54399 and they have worked well so far.
>I am getting new processors next month and hope to be able to reduce
>memory pressures in the test systems as I can give them a larger fish
>bowl to splash about in.
>
>Any more current guidance on setting MCCAFCTH?
>
One of our sysplex environments is like yours (5000,6000). I removed
those settings some time ago from the other environments after the
fix(es) for OW54399 came out (I wasn't supporting the one that still
has it at the time).
However, I remembered this being mentioned again at SHARE in Boston in
Kathy Walsh's performance hot topics as related to OA12185 which was
closed SUG (see below). So I guess you still may want to code it.
The OEM sort mentioned in the APAR is probably SYNCSORT, so if you
run SYNCSORT (I think you do, and so do we) you may want to check
with them about this issue.
APAR Identifier ...... OA12185 Last Changed ........ 05/10/18
IAXUA SPIN LOOP LARGE PGSER FIX PAGE FIX
ERROR DESCRIPTION:
An excessive spin loop is experienced in module Iaxua when a
very large pagefix is issued at a point where central storage
is constrained. This most often occurs within an application
that is running an oem sort where the sort has issued a page
fix for a multi-megabyte range of storage.
While it is documented under the PGSER macro that there is
potential for a spin loop to occur for page fixes over one
megabyte, this apar is being created explore a means to avoid
receiving the excessive spin MSGIEE178I (IEE178I) and
abend071.
LOCAL FIX:
In the ieaoptxx parmlib member specify, or change if already
specified, the mccafcth parameter to:
mccafcth=(4000,4500)
COMMENTS:
This APAR is being closed as SUG APAR. Development will
consider fixing it in a future release if the fix is
feasible.
Happy Holidays and Happy New Year Sam (and to all my friends
and fellow list members)!
Cheers,
Mark
--
Mark Zelden
Sr. Software and Systems Architect - z/OS Team Lead
Zurich North America and Farmers Insurance Group
mailto: mark.zelden@xxxxxxxxxxxx
Systems Programming expert at http://Search390.com/ateExperts/
Mark's MVS Utilities: http://home.flash.net/~mzelden/mvsutil.html
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to listserv@xxxxxxxxxxx with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
.
- Prev by Date: RE: Another incident
- Next by Date: D'oh !!!
- Previous by thread: D'oh !!!
- Next by thread: Re: Rule of thumb about paging / page fault rate
- Index(es):
Relevant Pages
|
