optimizedAcknowledge flag with performance issue

[zaoliu <za...@gmail.com>]

I set the optimizedAcknowledge flag in the consumer client to patch the
knowledge for boosting performance by following the link below: 
http://open.iona.com/wiki/display/ProdInfo/FUSE+Message+Broker+Performance+Tuning+Guide

By setting the optimized Acknowledge, the performance do greatly improved
from 14000 msgs/s to 20000 msgs/s. 

My testing enviroment has three machines: producer, consumer and broker are
all in different machine and I am using ActiveMQ 5.0. For get above best
result, I just use one producer and one consumer thread. Then I want to see
how the performance changed when I improve the number of consumer threads.
Since in my consumer code, the consumer does nothing except counting the
number of messages received when be notified, so adding number of consumers,
the performance may get down a little bit as my expectation. But the result
make me surprised, below is the result (all using one producer):
1 consumer:  ~20000/s
2 consumers:   ~18000/s
3 consumers:  ~1500/s
4 consuerms:  ~500/s

When I reset optimizedKnowledge flag, the result get back to normal:
1 consumer:  ~15000/s
2 consumers: ~14000/s
3 consumers:  ~12000/s
4 consumers   ~11000/s

How optimizedKnowlwege flag affect the performance for threads so much. Any
ideas?

Thanks,
Zao


-- 
View this message in context: http://www.nabble.com/optimizedAcknowledge-flag-with-performance-issue-tp15641642s2354p15641642.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.

Re: optimizedAcknowledge flag with performance issue

[David Sitsky <si...@nuix.com>]

Have you tried using different prefetch sizes for your queues?  I 
imagine in your application, this could make a big difference too.

Cheers,
David

zaoliu wrote:
> I set the optimizedAcknowledge flag in the consumer client to patch the
> knowledge for boosting performance by following the link below: 
> http://open.iona.com/wiki/display/ProdInfo/FUSE+Message+Broker+Performance+Tuning+Guide
> 
> By setting the optimized Acknowledge, the performance do greatly improved
> from 14000 msgs/s to 20000 msgs/s. 
> 
> My testing enviroment has three machines: producer, consumer and broker are
> all in different machine and I am using ActiveMQ 5.0. For get above best
> result, I just use one producer and one consumer thread. Then I want to see
> how the performance changed when I improve the number of consumer threads.
> Since in my consumer code, the consumer does nothing except counting the
> number of messages received when be notified, so adding number of consumers,
> the performance may get down a little bit as my expectation. But the result
> make me surprised, below is the result (all using one producer):
> 1 consumer:  ~20000/s
> 2 consumers:   ~18000/s
> 3 consumers:  ~1500/s
> 4 consuerms:  ~500/s
> 
> When I reset optimizedKnowledge flag, the result get back to normal:
> 1 consumer:  ~15000/s
> 2 consumers: ~14000/s
> 3 consumers:  ~12000/s
> 4 consumers   ~11000/s
> 
> How optimizedKnowlwege flag affect the performance for threads so much. Any
> ideas?
> 
> Thanks,
> Zao
> 
> 


-- 
Cheers,
David

Nuix Pty Ltd
Suite 79, 89 Jones St, Ultimo NSW 2007, Australia    Ph: +61 2 9280 0699
Web: http://www.nuix.com                            Fax: +61 2 9212 6902