| Author |
Message
|
| MacArKun |
 Posted: Mon Dec 18, 2006 10:51 pm Post subject: Every morning system is idle the transaction slow ~~ |
 |
|
Newbie
Joined: 18 Dec 2006
Posts: 4
|
AIX 5.3 MQ 5.3 CSD6
this pic is system transaction response time
Y:transaction response time(1000=1sec)
how can i do ~~ |
|
| Back to top |
|
 |
| Vitor |
| Posted: Tue Dec 19, 2006 12:14 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
There are multiple factors at work when it comes to performance, most of them not directly MQ. How have you established that MQ is the problem, as opposed to database issues, resource shortage, high CPU transactions, etc, etc, etc....?
Also "how do you" what? Improve performance? Improve throughput? Explain this to your boss? Effectively blame another area?
(Never discount the last, if only as a delaying tactic while you investigate) 
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| MacArKun |
| Posted: Tue Dec 19, 2006 1:18 am Post subject: |
|
|
Newbie
Joined: 18 Dec 2006
Posts: 4
|
| Vitor wrote: |
| There are multiple factors at work when it comes to performance, most of them not directly MQ. How have you established that MQ is the problem, as opposed to database issues, resource shortage, high CPU transactions, etc, etc, etc....? |
Server IBM pSeries 570 Model 9117
CPU 1.9GHz POWER4 x 16
Memory 64GB
DISK HITACHI TagmaStore Universal Storage Platform
Utilization
 |
|
| Back to top |
|
|
| Vitor |
| Posted: Tue Dec 19, 2006 1:36 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
So what you're saying is that response time grows as the system becomes more heavily loaded during peak hours?
And? 
At first glance I would say you have an application bottleneck as the resource usage seems constant but what further investigation have you already done (aside from these very well presented graphs)? What is the expected transaction performance? Has it degraded over time or always been bad? What is the target figure?
What are you asking?
("Help me" is all very well, but we need something more specific or you need an on-site consultant for 3 months!)
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| MacArKun |
| Posted: Tue Dec 19, 2006 1:45 am Post subject: |
|
|
Newbie
Joined: 18 Dec 2006
Posts: 4
|
According to application log, we found out that MQ has positive correlation with transaction volume in our environment. For instance, the average response time of each transaction at night takes 0.9 extra seconds when compared to transactions running at day. The difference is that our system usually has large transaction volume during daytime and less volume at night. MQ's performance follows that pattern too.
What could be the causes of that? What would be your advice to us?
Thanks a lot for the help. |
|
| Back to top |
|
|
| MacArKun |
| Posted: Tue Dec 19, 2006 1:53 am Post subject: |
|
|
Newbie
Joined: 18 Dec 2006
Posts: 4
|
| Vitor wrote: |
So what you're saying is that response time grows as the system becomes more heavily loaded during peak hours?
And?
At first glance I would say you have an application bottleneck as the resource usage seems constant but what further investigation have you already done (aside from these very well presented graphs)? What is the expected transaction performance? Has it degraded over time or always been bad? What is the target figure?
What are you asking?
("Help me" is all very well, but we need something more specific or you need an on-site consultant for 3 months!) |
We had local IBM consultants taken thorough investigation but without any good news. Our goal is simple. It is to achive constant transaction response time for the whole day.
Response time remains well satisfied except during night time. If we have to come out a figure, that would be under 0.5 seconds for each transaction. |
|
| Back to top |
|
|
| Michael Dag |
| Posted: Tue Dec 19, 2006 2:59 am Post subject: |
|
|
Jedi Knight
Joined: 13 Jun 2002
Posts: 2607
Location: The Netherlands (Amsterdam)
|
is this local MQ traffic or are there other queuemanagers via channels involved?
_________________
Michael
MQSystems Facebook page |
|
| Back to top |
|
|
| exerk |
| Posted: Tue Dec 19, 2006 3:15 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
I'll ask the obvious question...are you running back-ups in 'quiet hours'? Is there a difference in the server usage profile of day and night, i.e. what runs at night that doesn't run in the day?
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
| fjb_saper |
| Posted: Tue Dec 19, 2006 4:22 am Post subject: |
|
|
Grand High Poobah
Joined: 18 Nov 2003
Posts: 20756
Location: LI,NY
|
And how often do you get messages at night? If your channels are triggered and each message needs to trigger the channel you will get some added latency due to channel triggering... So negotiate a lower SLA for the first message.... 
_________________
MQ & Broker admin |
|
| Back to top |
|
|
| RogerLacroix |
| Posted: Tue Dec 19, 2006 8:58 am Post subject: |
|
|
Jedi Knight
Joined: 15 May 2001
Posts: 3264
Location: London, ON Canada
|
| MacArKun wrote: |
| CPU 1.9GHz POWER4 x 16 |
If I'm not mistaken, you are stating that the server has 16 CPUs?
But you are showing us a summary graph of all CPUs with the values averaged togther. Lets see the same day but by individual CPUs.
I've seen this before when an application is not multi-threaded and therefore, can only use 1 CPU. Hence, you end up with 15 CPUs at zero utilitization and 1 CPU at 100%.
A well written multi-threaded application should scream on a 16 CPU Power4 server.
Regards,
Roger Lacroix
_________________
Capitalware: Transforming tomorrow into today.
Connected to MQ!
Twitter |
|
| Back to top |
|
|
|
|
