Ruzi
"In each test, CPU usage was 03."
What type of performance impact are you looking for CPU usage or
throughput? I believe most of the answers that you've received address
throughput not CPU usage. They are improved response time due to less time
waiting on MQ to force the data to the queues.
Rick
|---------+--------------------------------------->
| | Ruzi R <[EMAIL PROTECTED]> |
| | |
| | Sent by: MQSeries List |
| | <[EMAIL PROTECTED]> |
| | |
| | |
| | |
| | Tuesday November 18, 2003 06:53 AM |
| | Please respond to MQSeries List |
| | |
|---------+--------------------------------------->
>----------------------------------------------------------------------------------------------------|
|
|
| To: [EMAIL PROTECTED]
|
| cc:
|
| Subject: Re: Impact of Syncpoint?
|
>----------------------------------------------------------------------------------------------------|
Thanks Scott...
I should have stated the scenario a bit more clearly.
Currently, an app is getting messages (Pers or
non-pers) outside UOW contrary to my recommendation. I
asked them to use syncpoint for persistent messages
for abvious reasons. Since they have to deal with Pers
and non-pers msgs in the same app, I suggested the use
of MQGMO_SYINCPOINT_IF_PERSISTENT option. And they
will have to commit after every Pers message. Now
they are asking me how much impact (postive or
negative) this syncpointing will have on performance.
To that end, I have just done the following 2 tests,
on the mainframe 5.3, which produced the same result
(i.e. no difference): I used 1K long 1000 Persistent
mesgs ineach case:
1- MQGET loop:
- Get with no-syncpoint
- Write the msg to a file
End-loop
- Commit (after the 1000th message)
2- MQGET loop:
- Get with syncpoint
- Write the msg to a file
- Commit
End-loop
In each test, CPU usage was 03. I wonder why your test
results are different from mine. Were your tests done
on NT? Even so, I did not think that there would be
this much difference. Any ideas?
Thanks,
Ruzi
Scott <[EMAIL PROTECTED]> wrote:
> You still get an improvement by using syncpoint
> after every message. The
> test I did reduced the time for 1000 persistent
> messages from 25 to 15
> seconds...
>
> Regards
> John Scott
> IBM Certified Specialist - MQSeries
> Argos Ltd
>
>
> -----Original Message-----
> From: Ruzi R [mailto:[EMAIL PROTECTED]
> Sent: 17 November 2003 19:03
> To: [EMAIL PROTECTED]
> Subject: Re: Impact of Syncpoint?
>
>
> Paul,
>
> Thanks for the response. But the requirement is that
> we have to do a commit after every persistent
> message
> read off the queue (as opposed to after a batch of
> messages read).
>
> Thanks,
>
> Ruzi
>
> --- Paul Clarke <[EMAIL PROTECTED]> wrote:
> > Depends what you mean by 'impact'. I assume you're
> > talking about persistent
> > messages here. Generally speaking the addition of
> > syncpoint to persistent
> > message operations will increase the performance
> > since we only need to
> > force the log at commit time rather than for every
> > message.
> >
> > Download my MA01 support pac if you want to have a
> > play.
> > Choose a local queue, say Q1
> > Load the queue with messages :-
> >
> > q -oQ1 -ap
> >
> > at the prompt type
> > #1000
> >
> > you now have 1000 persistent messages on your
> queue.
> >
> > Get them off by specifying :-
> >
> > q -IQ1 -t -q
> >
> > this will printout something like
> >
> > 1000 Interations in 441.00 ms, Average = 0.44ms or
> > 2267.6 per second
> >
> > This is doing your gets outside of syncpoint.
> > To do it inside of syncpoint you can tell queue
> how
> > often to commit, let's
> > say every 100 messages
> >
> > Load up your queue again and then type
> >
> > q -IQ1 -t -q -p100
> >
> > The line now reads
> > 1000 Interations in 90.00 ms, Average = 0.09ms or
> > 11111.1 per second
> >
> > By doing the gets under syncpoint (and quite a lot
> > of them) we're now about
> > 5 times faster.
> >
> > These were the numbers produced on my ThinkPad and
> > will depend greatly on
> > the speed of your disks, the type of disks, your
> CPU
> > speed etc etc.
> >
> > There are also, I believe, a number of support
> pacs
> > which detail the speed
> > of MQ operations on the support pac website.
> >
> > Hope this helps,
> > P.
> >
> > Paul G Clarke
> > WebSphere MQ Development
> > IBM Hursley
> >
> >
> >
> > |---------+---------------------------->
> > | | Ruzi R |
> > | | <[EMAIL PROTECTED]|
> > | | M> |
> > | | Sent by: MQSeries|
> > | | List |
> > | | <[EMAIL PROTECTED]|
> > | | N.AC.AT> |
> > | | |
> > | | |
> > | | 17/11/2003 16:59 |
> > | | Please respond to|
> > | | MQSeries List |
> > |---------+---------------------------->
> >
> >
>
>
---------------------------------------------------------------------------
> --------------------------------------------|
> > |
> >
> > |
> > | To: [EMAIL PROTECTED]
> >
> > |
> > | cc:
> >
> > |
> > | Subject: Impact of Syncpoint?
> >
> > |
> > |
> >
> > |
> > |
> >
> > |
> >
> >
>
>
---------------------------------------------------------------------------
> --------------------------------------------|
> >
> >
> >
> > MQ 5.3/W2K I have been asked to find out the
> > "performance impact" of using Syncpoint in MQGETs.
> > Is
> > there a document around that deals with this
> issue?
> >
> > Thanks,
> >
> > Ruzi
> >
> > Instructions for managing your mailing list
> > subscription are provided in
> > the Listserv General Users Guide available at
> http://www.lsoft.com
> > Archive: http://vm.akh-wien.ac.at/MQSeries.archive
> >
> > Instructions for managing your mailing list
> > subscription are provided in
> > the Listserv General Users Guide available at
> http://www.lsoft.com
> > Archive: http://vm.akh-wien.ac.at/MQSeries.archive
>
> Instructions for managing your mailing list
> subscription are provided in the
> Listserv General Users Guide available at
> http://www.lsoft.com
> Archive: http://vm.akh-wien.ac.at/MQSeries.archive
>
>
>
***********************************************************************************
>
> Click here to visit the Argos home page
> http://www.argos.co.uk
>
> The information contained in this message or any of
> its attachments may be privileged and confidential,
> and is intended exclusively for the addressee.
> The views expressed may not be official policy, but
> the personal views of the originator.
> If you are not the addressee, any disclosure,
> reproduction, dissemination or use of this
> communication is not authorised.
> If you have received this message in error, please
> advise the sender by using the reply facility in
> your e-mail software.
> All messages sent and received by Argos Ltd are
> monitored for virus, high risk file extensions, and
> inappropriate content. As a result users should be
> aware that mail maybe accessed.
>
> Instructions for managing your mailing list
> subscription are provided in
> the Listserv General Users Guide available at
> http://www.lsoft.com
> Archive: http://vm.akh-wien.ac.at/MQSeries.archive
Instructions for managing your mailing list subscription are provided in
the Listserv General Users Guide available at http://www.lsoft.com
Archive: http://vm.akh-wien.ac.at/MQSeries.archive
Instructions for managing your mailing list subscription are provided in
the Listserv General Users Guide available at http://www.lsoft.com
Archive: http://vm.akh-wien.ac.at/MQSeries.archive
