Paul/John: I remember the same item and if memory serves, they just
switched off the recording in the SMF PARMs. I'm a little rusty here but I
think the records are essentially built regardless and the only question is
do you want to record it or not. I may not quite have it right but I'm
sure
Shouldn't that be DED!!! Or maybe DOH!!!
-Original Message-
From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On
Behalf Of Jim Mulder
Sent: Tuesday, April 25, 2006 5:04 PM
To: IBM-MAIN@BAMA.UA.EDU
Subject: Re: Overhead of SMF Records
IBM Mainframe Discussion List IBM-MAIN
John S. Giltner, Jr. wrote:
I vaguly remember reading a paper (about 7 or 8 years ago) where
somebody turn off SMF recording and saw no measureable difference in CPU
utilziation.
I wonder how they measured it?
Regards, Paul Gillis
Paul Gillis [EMAIL PROTECTED] wrote in message news:[EMAIL PROTECTED]...
John S. Giltner, Jr. wrote:
I vaguly remember reading a paper (about 7 or 8 years ago) where
somebody turn off SMF recording and saw no measureable difference in CPU
utilziation.
I wonder how they measured it?
On Mon, 24 Apr 2006 00:00:00 GMT, Ted MacNEIL
[EMAIL PROTECTED] wrote:
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
In a message dated 4/25/2006 8:02:22 A.M. Central Daylight Time,
[EMAIL PROTECTED] writes:
Who knows how much collection costs?
There are three components of collection that cause overhead directly
attributable to SMF, but most of which is not measurable: (1) creation of
data
elements
-Original Message-
From: IBM Mainframe Discussion List
[mailto:[EMAIL PROTECTED] On Behalf Of John S. Giltner, Jr.
Sent: Monday, April 24, 2006 10:08 PM
To: IBM-MAIN@BAMA.UA.EDU
Subject: Re: Overhead of SMF Records
I vaguly remember reading a paper (about 7 or 8 years ago
In a message dated 4/25/2006 2:42:56 A.M. Central Standard Time,
[EMAIL PROTECTED] writes:
wonder how they measured it?
With an hour-glass?
--
For IBM-MAIN subscribe / signoff / archive access instructions,
send email
Are you in a large shop now?
Approx 3600 MIPS.
How busy are your processors? In our
shop (where we like to run all the processors close to 100% all day),
We try, but our business is cyclic.
putting SMF dumps at the bottom of the totem pole would never work. You have
to be able to dump
IBM Mainframe Discussion List IBM-MAIN@BAMA.UA.EDU wrote on 04/25/2006
09:24:37 AM:
IBM has some undocumented monitor call
instructions inside the I/O first level interrupt handler for
instrumentation
purposes. Maybe they also have some way to measure the overhead of
building an
For what it's worth, and I'm sure this can start a great debate, I allow the
SMF
processes to take what they need to get in, finish what they need to, and
get out.
I have seen customer sites move SMF out of SYSTEM, and have MANx Dumps run
at lower importance. I generally recommend giving it all
IBM Mainframe Discussion List IBM-MAIN@BAMA.UA.EDU wrote on 04/25/2006
11:30:56 AM:
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.
I had never
I guess some IBM developers have a sense of humor after all.
I always thought they did.
Remember the prefix for SMS is IGD, and the main programme is called IGDZILLA.
I, Godzilla
-
-teD
O-KAY! BLUE! JAYS!
Let's PLAY! BALL!
Jim Mulder wrote:
[D0D wait state code for the DOD]
I guess some IBM developers have a sense of humor after all.
Any company that allows developers to ship macros named $DILBERT and
$DOGBERT (JES2), and a module named GIMISEX (SMP/E) is surely not without a
sense of humour. And then from the
IIRC the continued to cut RMF records.
Paul Gillis wrote:
John S. Giltner, Jr. wrote:
I vaguly remember reading a paper (about 7 or 8 years ago) where
somebody turn off SMF recording and saw no measureable difference in
CPU utilziation.
I wonder how they measured it?
Regards, Paul
At 00:00 + on 04/25/2006, Ted MacNEIL wrote about Re: Overhead of
SMF Records:
I always thought they did.
Remember the prefix for SMS is IGD, and the main programme is called IGDZILLA.
I, Godzilla
If SMS's prefix was IMG, they could have named the main module as
IMGOJIRA (I'm GOJIRA
Please do not laugh.
My management is asking me how much overhead is consumed by z/OS for
processing the SMF records. I have no idea. Has anyone ever researched this
and can share some info on it.
Thanks in advance.
Joel Wolpert
Director - Performance and Capacity Planning
Shared Data Center
In a message dated 4/24/2006 3:54:05 P.M. Central Standard Time,
[EMAIL PROTECTED] writes:
My management is asking me how much overhead is consumed by z/OS for
processing the SMF records. I have no idea. Has anyone ever researched this
and can share some info on it.
PHB's say the
My management is asking me how much overhead is consumed by z/OS for
processing the SMF records.
Producing SMF records is the cost.
Presenting them to SMF and having the task write them out is cheap.
Because it's asynchronous.
Move SMF down in importance and you won't suffer.
The records will
Thanks. This is good info. Does this mean that if I measure the CPU usage
of the SMF address space I will be able to approximate the CPU overhead?
In a message dated 4/24/2006 4:01:48 P.M. Central Daylight Time,
[EMAIL PROTECTED] writes:
My management is asking me how much overhead is consumed by z/OS for
processing the SMF records. I have no idea. Has anyone ever researched this
and can share some info on it.
The eternal correct
In a message dated 4/24/2006 4:12:25 P.M. Central Daylight Time,
[EMAIL PROTECTED] writes:
Does this mean that if I measure the CPU usage
of the SMF address space I will be able to approximate the CPU overhead?
No. The greatest part of the overhead is in creating them, and this takes
Are they asking about the overhead of collecting them, and/or the overhead
of processing them to
derive useful information - or hadn't they thought of that.
Please do not laugh.
My management is asking me how much overhead is consumed by z/OS for
processing the SMF records. I have no idea.
They are asking about the overhead of collecting them. I know how to
measure the overhead of processing them.
Thanks.
Does this mean that if I measure the CPU usage
of the SMF address space I will be able to approximate the CPU overhead?
No!
Most of the overhead is in the sub-systems producing the records.
CMF, DB2, NPM, SILO, RMF, JES2, etc.
SMF actually writing them is cheap.
-
-teD
O-KAY! BLUE! JAYS!
Please do not laugh.
Ok. My lips are contorted into a strange shape like a chortle in
progress, but I swear I am not laughing. Well, ok, maybe just a little.
My management is asking me how much overhead is consumed by z/OS for
processing the SMF records. I have no idea. Has anyone ever
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
O-KAY! BLUE!
I vaguly remember reading a paper (about 7 or 8 years ago) where
somebody turn off SMF recording and saw no measureable difference in CPU
utilziation. As other have said the overhead is in collecting
infromation needed to create the record and creating the record. Some
system will do
28 matches
Mail list logo
