| Author |
Message
|
| sebastia |
 Posted: Sat Feb 28, 2009 2:02 am Post subject: mq v7 performance figures |
 |
|
Grand Master
Joined: 07 Oct 2004
Posts: 1003
|
Good morning.
I have a customer that asked me 2 very simple questions :
1) using mq v7 and a specific wintel machine
(QuadCore, 4GB RAM, 3GHz CPU)
will it be able to manage 2 million messages a day,
being each one 32 KB of size ?
2) what wintel machine specs do we need
to manage 4 million messages a day (32KB each) ?
Yes, I do have "MP7I" ... thanks. Sebas. |
|
| Back to top |
|
 |
| ramires |
| Posted: Sat Feb 28, 2009 4:13 pm Post subject: |
|
|
Knight
Joined: 24 Jun 2001
Posts: 523
Location: Portugal - Lisboa
|
Sebastia,
4.000.000 a day and if a day is 122.400 seconds this makes 47 message a second. One important thing to consider is message persistence.
But 47 msg/sec seems achievable.
Of course there are a other parameters to consider.
Regards
joao |
|
| Back to top |
|
|
| sebastia |
| Posted: Sat Feb 28, 2009 4:53 pm Post subject: |
|
|
Grand Master
Joined: 07 Oct 2004
Posts: 1003
|
Thanks, Joao.
But my customer needs something more specific than this.
MP7I says ...
a) an xSeries 365 4 cpu 3.0ghz Intel xeon with 4Gb of RAM
b) figure 24 at MP7I indicates a 32 KB message is moved by 0,15 msg,
so MQ v7 can move 1000/0,15 = 6.666 msg per second.
So, 47 msg/sec is not the problem.
But ... how can I prove that ?
a) creating a scenario ?
b) having a sample customer
c) ???
Thanks. |
|
| Back to top |
|
|
| ramires |
| Posted: Sat Feb 28, 2009 5:47 pm Post subject: |
|
|
Knight
Joined: 24 Jun 2001
Posts: 523
Location: Portugal - Lisboa
|
Maybe with a scenario.
I've just did a test with MO71 to write 100.000 msg with 32.000 bytes length to a local q and the result was 309 msg/sec. I'm with 6.0.2.5
Regards |
|
| Back to top |
|
|
| standa |
| Posted: Mon Mar 02, 2009 3:52 am Post subject: msg allocation per hour in the day |
|
|
Novice
Joined: 08 Sep 2006
Posts: 10
Location: CZECHREP
|
Consider also msg. rates per hour in day.
Situation where there is steady flow is quite different to the situation where half of the msgs are sent in 2hours and rest in the rest of the day.
_________________
standa. |
|
| Back to top |
|
|
| sebastia |
| Posted: Mon Mar 02, 2009 3:57 am Post subject: |
|
|
Grand Master
Joined: 07 Oct 2004
Posts: 1003
|
Thanks, mr STANDA - I agree the "peaks" are important
but my customer does not have them,
as Request/Reply is used,
this is "On-Line" transactions are the MQ Clients,
and MQ at Host is the Server.
There are no BATCH transfers at midnite ... yet |
|
| Back to top |
|
|
| gbaddeley |
| Posted: Mon Mar 02, 2009 6:15 pm Post subject: |
|
|
Jedi Knight
Joined: 25 Mar 2003
Posts: 2538
Location: Melbourne, Australia
|
Network capacity could also be an issue. You really need to conduct a proof of concept on a network that has a similar work load.
There may also be capacity limitations in backend and downstream applications, particularly if they use databases and are running on remote systems.
Provided the h/w and s/w is up to spec, generally MQ performance is very good and the limiting factors will be found elsewhere.
_________________
Glenn |
|
| Back to top |
|
|
| sebastia |
| Posted: Tue Mar 03, 2009 2:52 am Post subject: |
|
|
Grand Master
Joined: 07 Oct 2004
Posts: 1003
|
mr GBAD - I agree completely.
Network or Disk shall be the bottlenecks
Thanks. |
|
| Back to top |
|
|
|
|
