Hi Ted,
The SMF address space is classified SYSTEM.
Here I have our SMFDMPxx jobs used to dump the MANx data set as an IMP 1
workload.
To stay ahead of SMF data volumes and prevent data loss especially at
midnight some things I have done here include:
Leveraged z/OS R6 feature to customize SMF buffers through SMFPRM00 to
allow use of up to 640M of SMF buffers 5 times the IBM default. This
does not use real storage when the buffers are not in use which is an
optimization IBM made the way SMF handles data buffers in z/OS R6.
Leveraged z/OS R6 feature to set a threshold for warning to suppress
spurious warning messages for buffer utilization now 50% before IEE986E
is issued.
Implemented a 4th SMF MAN data set so that all systems now have
additional recording time if dumps are delayed
SMF dump jobs have been prioritized into a service class for critical
batch
Previously we had made the following enhancements which reduced the run
time of our production SMF daily midnight job to 1/4 of it's prior
16x data striped output data sets from MANx SMFDMP jobs to significantly
improve nightly throughput of daily consolidation jobs
DFSMS compression for SMF data to improve I/O efficiency and reduce
required space
Thruput Manager gateway SMF reader/writer binding resource to prevent
contention with daily consolidation jobs
Daily consolidation jobs had been prioritized into a service class for
critical batch
Optimized SMF MAN data set VSAM CISIZE from 4K to 26K
I cannot afford to dink with the WLM classification of the SMF address
space or to allow dump processing to be slowed down or stalled running
in discretionary. SMF dumps occur when we may be running at 100%
utilization.
SMF is one of the unique strengths of the zSeries platforms we only wish
we had this kind of rich, uniform data stream on AIX, Linux, Windows.
SMF is also a key part of fulfilling our auditing and SOX compliance
requirements.
We are not quite at the level of kill the system rather than run without
an SMF audit trail ("wait state D0D") but we address any SMF data loss
with steps taken to prevent it from recurring.
Best Regards,
Sam Knutson, GEICO
Performance and Availability Management
mailto:[EMAIL PROTECTED]
(office) 301.986.3574
"Think big, act bold, start simple, grow fast..."
-----Original Message-----
From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On
Behalf Of Ted MacNEIL
Sent: Sunday, April 23, 2006 8:00 PM
To: IBM-MAIN@BAMA.UA.EDU
Subject: Re: Overhead of SMF Records
>Don't forget to add in the cost of the periodic IFASMDFP jobs that copy
the SMF data from the SYS1.MANx datasets
I make those IMP=5.
The cost is nothing, since I bought the machine for other work.
IFASMFDP fills in the valleys that would not be used if they weren't
there!
-
-teD
====================
This email/fax message is for the sole use of the intended
recipient(s) and may contain confidential and privileged information.
Any unauthorized review, use, disclosure or distribution of this
email/fax is prohibited. If you are not the intended recipient, please
destroy all paper and electronic copies of the original message.
----------------------------------------------------------------------
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
