While we're on the topic of WLM: Remember this topic back in 2007? And
specifically *this*:
Exactly my point: I think *someone* needs to take a real hard look at how
this product was ported to z/OS. (I didn't mention that it was ported from
the 'open world', did I?) I also think to 'play nice'
I had the idea of getting automation to listen
to IEF404I BPXAS - ENDED - TIME=00.31.51
$HASP395 BPXASENDED
(one of the two), and whenever it happens do the forks under the
assumption that that will not hinder the *actual* work that needs it.
Wouldn't is be sufficient to run
Barbara Nitz [EMAIL PROTECTED] wrote in message
news:[EMAIL PROTECTED]...
First of all, my apologies to the original poster that I had
appropriated his thread! I am finally changing the name
I have checked with our resident unix people - a little program to do
fork() will be easy to
Be careful: it was some time ago when I had my USS course, but I
remember that there are fork()'s that can remain within the same
address space and fork()'s that require a new (BPXAS) address space.
There are *processes* that can share an address space. If a new
process is create with fork()
Hunkeler Peter , KIUK 3 [EMAIL PROTECTED] wrote in
message
news:[EMAIL PROTECTED].
...
Be careful: it was some time ago when I had my USS course, but I
remember that there are fork()'s that can remain within the same
address space and fork()'s that require a new (BPXAS) address space.
First of all, my apologies to the original poster that I had appropriated his
thread! I am finally changing the name
snipping Peters good stuff about UNIX
I did check into syslog. In general, we have quite a few BPXASs around, so when
a burst of work comes in (messages uji001 written
Or are we really now talking about outSTUPIDing WLM? :-) :-)
Cheers, Martin
Martin Packer
Performance Consultant
IBM United Kingdom Ltd
+44-20-8832-5167
+44-7802-245-584
[EMAIL PROTECTED]
Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with
Barbara,
Here's one trick I've used in the past when working in this area is this:
go into SDSF, do a DA ALL, and SORT it by ASID. You should then see some
number of idle address spaces, named BPXAS. Then when a process forks and
needs to do something, you'll see that BPXAS change it's name
Hunkeler Peter (KIUK 3) wrote:
I didn't mean to ask for the quantity but for how zIIPs can help that
software, i.e. is it zIIP eligible work. Anyway zIIPs can drop software
cost but would not make the system run faster, would they?
They can. But, only if the processor is a capacity model
Hello,
nice thread - but the heading isn't quite right.
These processes play by WLM rules: pop up all by the same time,use little cpu
and go to sleep again. RMF will show bursts of activity and
long periods with low activity. I've seen this with batch jobs
having common i/o patterns and low cpu
Have to run the zIIPalyzer to make an educated guess. Maybe a little,
maybe
enough to make it thru the humps. W/O measurement just idle
speculation.
I didn't mean to ask for the quantity but for how zIIPs can help that
software, i.e. is it zIIP eligible work. Anyway zIIPs can drop software
that don't play by
WLM's rules
as per previous post ... there was the vs/pascal implementation ported
from vm ... with a diagnose instruction simulation done in os ... and
then the vtam-based implementation (that started out only being
correct if it had lower thruput than lu6.2).
some part
by: IBM Mainframe Discussion List IBM-MAIN@BAMA.UA.EDU
10/19/2007 01:54 AM
Please respond to
IBM Mainframe Discussion List IBM-MAIN@BAMA.UA.EDU
To
IBM-MAIN@BAMA.UA.EDU
cc
Subject
Re: Are there tasks that don't play by WLM's rules
Thanks to all who responded. I find it really interesting
The really big roll-out was at 2.5, and it, too, was a big porting effort
because a number of the redesigned OS/390 processes were Unix processes
ported to Unix System services.
I remember now (surprise!).
We did not move STK's VTS to TCP/IP from coax until 2.7, even though there was
a lot of
In a message dated 10/19/2007 9:46:49 A.M. Central Daylight Time,
[EMAIL PROTECTED] writes:
software, i.e. is it zIIP eligible work. Anyway zIIPs can drop software
cost but would not make the system run faster, would they?
Depends on how much it offloads? Sort of like the supermarket
In a message dated 10/19/2007 3:28:05 P.M. Central Daylight Time,
[EMAIL PROTECTED] writes:
All in all, I would not hold up the TCP/IP reimplementation as a model
of a successful replacement of a poorly ported product. It took a
number releases before I would call it successful.
It
On Fri, 19 Oct 2007 09:38:02 +, Ted MacNEIL
[EMAIL PROTECTED] wrote:
When porting big applications or servers to z/OS UNIX the ...
It's not just z/OS UNIX.
The first implementation of TCP/IP on OS/390 was a port from VM.
And, it was a pig until they decided to re-implement by starting from
All work that can be dispatched on a zIIP will run at full speed
regardless of the speed of the general CPs. So, yes, a zIIP can
actually make work run faster.
My fault. I just didn't think of throttled engines.
--
Peter Hunkeler
Credit Suisse
-Original Message-
From: IBM Mainframe Discussion List
[mailto:[EMAIL PROTECTED] On Behalf Of Hunkeler Peter (KIUK 3)
Sent: Friday, October 19, 2007 9:45 AM
To: IBM-MAIN@BAMA.UA.EDU
Subject: Re: Are there tasks that don't play by WLM's rules
Have to run the zIIPalyzer to make
In a message dated 10/19/2007 1:31:46 A.M. Central Daylight Time,
[EMAIL PROTECTED] writes:
How would zIIPs help here?
Have to run the zIIPalyzer to make an educated guess. Maybe a little, maybe
enough to make it thru the humps. W/O measurement just idle speculation.
Same for some of
Re: Are there tasks that don't play by WLM's rules
The following message is a courtesy copy of an article
that has been posted to bit.listserv.ibm-main,alt.folklore.computers as
well.
[EMAIL PROTECTED] (Ted MacNEIL) writes:
It's not just z/OS UNIX.
The first implementation of TCP/IP on OS
Well, that's what I am attempting to do. Did I mention that the
vendor is IBM? And that the product is developed in the same
location as WLM? In Germany, no less?
Well, unfortunately, there are still too many people out there in
software development that don't have a clue what z/OS UNIX really
I didn't mean to ask for the quantity but for how zIIPs can
help that software, i.e. is it zIIP eligible work. Anyway
zIIPs can drop software cost but would not make the system
run faster, would they?
Possibly -- IIRC the specialty engines always run at full throttle,
regardless
-Original Message-
From: IBM Mainframe Discussion List On Behalf Of Hunkeler Peter (KIUK
3)
[ snip ]
I didn't mean to ask for the quantity but for how zIIPs can
help that software, i.e. is it zIIP eligible work. Anyway
zIIPs can drop software cost but would not make the system
The following message is a courtesy copy of an article
that has been posted to bit.listserv.ibm-main,alt.folklore.computers as well.
[EMAIL PROTECTED] (Ted MacNEIL) writes:
It's not just z/OS UNIX.
The first implementation of TCP/IP on OS/390 was a port from VM.
And, it was a pig until they
When porting big applications or servers to z/OS UNIX the developers must be
willing to change the design here and there to cope with z/OS UNIX and MVS
behaviour. This makes the difference between a
successful port and a not so successful one.
It's not just z/OS UNIX.
The first implementation
Well, zIIPs are cheaper than CPUs, but in this case maybe CuOD is the
best choice. If you pays more, then probably under powered anyway...
How would zIIPs help here?
--
Peter Hunkeler
Credit Suisse
--
For IBM-MAIN subscribe /
Barbara Nitz [EMAIL PROTECTED] wrote in message
news:[EMAIL PROTECTED]...
Mark and Shane,
Do you have enough engines and capacity where bursts won't hurt other
workloads? If so, I might be inclined to run this work in SYSSTC (I
assume discretionary won't work for this since you mentioned on
Kees,
- Is the WLM PI number the real problem, or is the performance actually
bad? A bad report about a well running application is not the end of the
world.
We started looking at this when we had complaints about missing throughput and
bad response times. Before that, we didn't even realize
Barbara Nitz [EMAIL PROTECTED] wrote in message
news:[EMAIL PROTECTED]...
Kees,
- Is the WLM PI number the real problem, or is the performance
actually
bad? A bad report about a well running application is not the end of
the
world.
We started looking at this when we had complaints about
Kees,
- How did this thing run in the 'open' world (or didn't it)? Machines
with dozens of processors to execute these 65 tasks immediately are not
common there either AFAIK.
No idea. The architecture in the 'open' world may be completely different.
- I am a little worried about your remark
Barbara Nitz [EMAIL PROTECTED] wrote in message
news:[EMAIL PROTECTED]...
The MVS implementation of BPX 'address space creation' was changed
back in the nineties to use RSM shared pages (a virtual overlay of the
creator's address space) because most of the parent address space pages
were
Kees,
thanks a lot for taking the time to discuss this with me! Now that you mention
it, it appears so obvious, and I was just too slow to see it. (Guess
somewhere along the way I got tangeld up in other tangents.)
I'll talk to my colleague and will try to implement it as soon as possible
On Oct 18, 2007, at 6:00 AM, Vernooy, C.P. - SPLXM wrote:
Barbara Nitz [EMAIL PROTECTED] wrote in message
news:[EMAIL PROTECTED]...
The MVS implementation of BPX 'address space creation' was changed
back in the nineties to use RSM shared pages (a virtual overlay of the
creator's address
On Thu, 18 Oct 2007 11:00:45 +0200, Barbara Nitz [EMAIL PROTECTED] wrote:
My biggest concern with the high PI is actually that WLM will only try to
help the service class every 3rd trip through WLM (so I was told).
Exactly why I think SYSSTC is the best place for this. It sounds like you
- I am a little worried about your remark about the (short) life of
the
address spaces: creating a new thread in Unix is just simply adding an
entry to a process table; if it requires address space creation in
z/OS,
this is a heavy overhead part added to kicking off the task. Can you
play around
Peter Hunkeler writes some good stuff about z/OS Unix;
PMFJI, but the last few remarks about being ported from open world
made me think. So here are a few thoughts related to z/OS UNIX:
details of BPXAS behaviors snipped
Exactly. Execution velocity as a goal mechanism has been a really bad
Date:Wed, 17 Oct 2007 10:46:45 +0200
From:Vernooy, C.P. - SPLXM [EMAIL PROTECTED]
Subject: Re: Are there tasks that don't play by WLM's rules
The problem here is, that WLM is not capable of managing its children
when they are a little more exotic than the usual kids in class. or at
least
In a message dated 10/18/2007 12:40:25 P.M. Central Daylight Time,
[EMAIL PROTECTED] writes:
Then get your boss to go scream at the vendor to put proper
instrumentation into their product. There's not a lot else you can do
Well, zIIPs are cheaper than CPUs, but in this case maybe CuOD is
Thanks to all who responded. I find it really interesting that just about
everyone agrees to put these things into sysstc, an idea that got vehemently
vetoed by the 'head WLM developer' when I suggested it in a lengthy conference
call.
Thanks to Peter Hunkeler for the good stuff about Unix. I
Hi Tom,
thanks for your explanation. That was basically what I already knew and what
WLM development told me in 5000 words or more.
That product not only sets several timers per address space, they also pop all
at the same time. (I looked at a system trace table of 1.5s length, and for
this
Barbara Nitz [EMAIL PROTECTED] wrote in message
news:[EMAIL PROTECTED]...
Hi Tom,
thanks for your explanation. That was basically what I already knew
and what WLM development told me in 5000 words or more.
That product not only sets several timers per address space, they also
pop all at the
Und Barbara schriebt:
And then comes Martin and makes me get out the dictionary! :-)
And as I always do whenever in a German-speaking country I must apologise
for my English. :-)
Right before saying and just because you're speaking in German doesn't
mean it's encrypted. :-)
MFG, Martin
I see the problem and the consequences, but I think your conclusion and
originally already the title are not and cannot be correct: it shouldn't
matter to WLM whether tasks want to play by WLM's rules or not. WLM
should manage the tasks, whether they want to play along or not. Where
is the rule
On Wed, 17 Oct 2007 10:28:06 +0200, Barbara Nitz [EMAIL PROTECTED] wrote:
Fact is that I am stuck with a very bad PI (and I cannot really do anything
about it in WLM), and when the fines are due to be paid (because we didn't
deliver on time) or the customer complaints come in, it's my head that
On Wed, 2007-10-17 at 07:42 -0500, Mark Zelden wrote:
Do you have enough engines and capacity where bursts won't hurt other
workloads? If so, I might be inclined to run this work in SYSSTC (I assume
discretionary won't work for this since you mentioned on time) and then
WLM doesn't have to
Mark and Shane,
Do you have enough engines and capacity where bursts won't hurt other
workloads? If so, I might be inclined to run this work in SYSSTC (I
assume discretionary won't work for this since you mentioned on time)
and then WLM doesn't have to manage it.
thanks for that suggestion. It
Date:Mon, 15 Oct 2007 10:57:09 +0200
From:Barbara Nitz [EMAIL PROTECTED]
Subject: Re: Are there tasks that don't play by WLM's rules
We run one of those newfangled things called WBIFN (MQS-Merva-Bridge to
swiftnet). That productive service class runs IMP1 exvel 40% (I believe
Hi Tom!
Tom wrote:
I have seen this same effect with a Peoplesoft
customer trying to use velocity for the DDF enclaves.
Tom, I'd be encouraging Peoplesoft customers to use CMTSTAT=INACTIVE (A
DB2 ZPARM) and the usual panoply of Multiperiod Response Time / Percentile
goals.
But this is all
that don't play by WLM's rules
I am cross posting this to the IBM-Main and the MXG listservs.
We are about to set the soft cap on our systems. I have been
informed
by my system programmers that in the past they have seen tasks that
don't play by WLM's rules when it comes to resource
Has anyone else experienced this with Direct-Connect or any other
application?
NDM is a cpu hog if you let it. We run it in discretionary, but then we do have
a lot classified to discretionary service classes (it's the default for all
STCs and all batch). I believe that most of NDMs cpu is
-Original Message-
From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On
Behalf Of Barbara Nitz
Sent: Monday, October 15, 2007 3:57 AM
To: IBM-MAIN@BAMA.UA.EDU
Subject: Re: Are there tasks that don't play by WLM's rules
Has anyone else experienced this with Direct-Connect
The reason was that there was a single job running in the test service
class and many jobs
running in production. In this case, the PI of the test service class
was greater than the PI of the production one, so WLM gave CPU cycles to
the single test job.
This wouldn't make any difference unless
snip
Has anyone else experienced this with Direct-Connect or any other
application? That is have they experienced the application taking over
the system even though it has a low importance level? This is important
to us as we want our CICS processing to be one of the last things to be
hurt if we
-Original Message-
From: IBM Mainframe Discussion List
[mailto:[EMAIL PROTECTED] On Behalf Of Kelman, Tom
Sent: Friday, October 12, 2007 1:43 PM
To: IBM-MAIN@BAMA.UA.EDU
Subject: Re: Are there tasks that don't play by WLM's rules
snip
John,
Do you have the SERVERS service
.
Sent: Friday, October 12, 2007 1:00 PM by John McKown
In a sense, yes. I've had test batch perform better than production
batch, despite a lower velocity and importance. The reason was that
there was a single job running in the test service class and many jobs
running in production. In
I am cross posting this to the IBM-Main and the MXG listservs.
We are about to set the soft cap on our systems. I have been informed
by my system programmers that in the past they have seen tasks that
don't play by WLM's rules when it comes to resource utilization. One is
the Direct-Connect
-Original Message-
From: IBM Mainframe Discussion List
[mailto:[EMAIL PROTECTED] On Behalf Of Kelman, Tom
Sent: Friday, October 12, 2007 12:04 PM
To: IBM-MAIN@BAMA.UA.EDU
Subject: Are there tasks that don't play by WLM's rules
I am cross posting this to the IBM-Main and the MXG
On Fri, 12 Oct 2007 13:48:13 -0500, McKown, John wrote:
I only use transaction response time for TSO users. Everything else,
including CICS, runs with simple velocity and importance service
classes. It works well enough for us.
--
I went the other way. All of our CICS's run with a transaction
On Fri, 2007-10-12 at 12:04 -0500, Kelman, Tom wrote:
We are about to set the soft cap on our systems. I have been informed
by my system programmers that in the past they have seen tasks that
don't play by WLM's rules when it comes to resource utilization. One is
the Direct-Connect (NDM
If you are running to transaction goals, your imp/velocity is irrelevant.
Importance is still relevent to transaction goals.
I think you know that, but your wording doesn't imply that.
-
Too busy driving to stop for gas!
--
For
61 matches
Mail list logo