| Author |
Message
|
| jefflowrey |
 Posted: Wed May 10, 2006 12:05 pm Post subject: |
 |
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
No, sorry. I knew that. I just lost track for a moment. 
Regardless, you should monitor the enqueue rate and dequeue rate, and see how it matches up.
There is still an MCA in the picture, but it's the client MCA that is executing the MQI calls on behalf of the client app.
You should also look at network utilization on the windows machine, to see if you're bound at the network layer.
_________________
I am *not* the model of the modern major general.
Last edited by jefflowrey on Wed May 10, 2006 1:38 pm; edited 1 time in total |
|
| Back to top |
|
 |
| PeterPotkay |
| Posted: Wed May 10, 2006 1:30 pm Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
What does the getting client do? If it inserts to a DB, that insert is gonna take x milliseconds. If you have 20 instances of clients trying to right to the DB, then they are all in contention for that one DB, and so they wait in line to do inserts. Maybe that's why you don't see significant improvement.
Or maybe you run all 20 client instances on the same server, and they are in contention for disk I/O or CPU.
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| vincent |
| Posted: Wed May 10, 2006 3:08 pm Post subject: |
|
|
Voyager
Joined: 21 Apr 2006
Posts: 81
|
No database connection. It is a simple message,2k size, just has to flow through from solaris to windows. but need to get 1500 messages per second.
-vincent |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Wed May 10, 2006 3:16 pm Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
How fast is your network connection between the two machines?
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| vincent |
| Posted: Wed May 10, 2006 3:46 pm Post subject: |
|
|
Voyager
Joined: 21 Apr 2006
Posts: 81
|
I did not measure the network. The workaround i have is to create the test application on solaris box, So both the put and get client application is created in solaris box, Still no luck. It is taking 33 seconds to process 2000 messages. It is approximatly 60 seconds. Within the 60 seconds, the reader client did not pull all messages. If i put more readers, it is slowing down the total performance.
-vincent. |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Wed May 10, 2006 3:50 pm Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
| PeterPotkay wrote: |
| What does the getting client do? ...... Or maybe you run all 20 client instances on the same server, and they are in contention for disk I/O or CPU. |
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| vincent |
| Posted: Wed May 10, 2006 5:56 pm Post subject: |
|
|
Voyager
Joined: 21 Apr 2006
Posts: 81
|
It is good assumption and may be true but i have put all the 20 instances on the same box and it was increased up to 10 times than one instance while adding messages
It is really behaving very bad while i retrieved. Litterllay there is no difference between single and multiple clients
-vincent. |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Wed May 10, 2006 8:07 pm Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20757
Location: LI,NY
|
Can you create a test in which your pulling app does nothing with the message but discard it? Do you still see the same rates? If the pulling rates have vastly improved, it is clearly the fault of the process in the pulling app.
If the pulling rate stays low it may be PMR time.
One other consideration, how long after the putting app has started do you run the pulling app? The queue depth could have a very dammaging influence if you use selectors or have to browse. When you go for speed the pulling app should do nothing but pull the first available message from the queue and process it and keep processing untill no more messages are available ... (get with wait 5 secs)...
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| nathanw |
| Posted: Thu May 11, 2006 2:10 am Post subject: |
|
|
Knight
Joined: 14 Jul 2004
Posts: 550
|
Am curious does the get application start at a high rate and then tail off or is it at a consistent rate all the time?
Could this be an environment / CPU / Memory problem rather than a Application MQ problem |
|
| Back to top |
|
|
| vincent |
| Posted: Thu May 11, 2006 4:01 am Post subject: |
|
|
Voyager
Joined: 21 Apr 2006
Posts: 81
|
saper,
That is what i am doing in my test applicaiton, pulling out and discarding them. I am still seeing the same rate. There is no improvement in pulling rate.
filled with 40000 messages and waiting for 5 minutes and pulling out all the 40000 by repeated Gets but the rate is very slow.
nathan,
The rate is consistent all the times. I have 2GB and 2CPU and the occupied percentage was 21% while MQ running and 1.5GM space was free.
-vincent |
|
| Back to top |
|
|
| mvic |
| Posted: Thu May 11, 2006 4:41 am Post subject: |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
| vincent wrote: |
saper,
That is what i am doing in my test applicaiton, pulling out and discarding them. I am still seeing the same rate. There is no improvement in pulling rate. |
I suggested this in one of your other threads, but didn't notice any feedback. I'll mention it again, as it seems the best approach to me: take a trace, and find out where the significant time is being taken. Concentrate on this thing.
Hope this helps |
|
| Back to top |
|
|
| vincent |
| Posted: Thu May 11, 2006 6:59 am Post subject: |
|
|
Voyager
Joined: 21 Apr 2006
Posts: 81
|
Will adding more reader pull out more messages or will it be same?
If it is same, is there any problem or is that how MQ behaves?
In contrast, Addding more provider actually increase the pump in rate?
I enabled trace, and did not provide me much delay and not able to make out any problem there
-vincent |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Thu May 11, 2006 7:08 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
Adding more readers will allow you to pull out more messages, assuming that you have not reached limits outside of MQ's control - like file i/o speed and network speed and cpu utilization and etc.
And assuming that your readers aren't interfering with each other.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| vincent |
| Posted: Thu May 11, 2006 7:30 am Post subject: |
|
|
Voyager
Joined: 21 Apr 2006
Posts: 81
|
Jeff, All readers reads from same Q at the same time. Each readers started speratly from each shell window. What do you mean readers are not inerfering withe each other? How do i see them if it interfering each other?
Trace is not providing any clue about this interference |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Thu May 11, 2006 7:37 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
I meant interefering by contending for cpu or network or memory resources that are otherwise scarce.
Two applications with an input handle on the same queue will never interfer with each other trying to access to the same message.
mvic is suggesting that you examine each of the steps listed in the traces, and see which ones have a longer gap in the reported time from the previous step.
In particular, something like this. Stop all the clients. enable client tracing. Start one client. have it pull all the messages. start another client. Put the 40,000 messages again. Then compare how long each application took for each step, and compare the same client app to see if that takes longer when another client is running.
I've forgotten - you're doing this on v5.3, right? Are you on the same CSD level on the server as the client? Are you on a *recent* CSD level on both the client and the server?
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
|
|
