activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From cobrien <>
Subject Re: question on ActiveMQ performance using 10GBP network card
Date Thu, 22 Jul 2010 17:15:19 GMT

There are several tuning params that can dramatically change the throughput. 
For starters I am wondering what your prefetch size is? do you use
transactions?   Can you provide your activemq.xml and let us know the size
of the messages and average and worse case  time required for a consumer to
process a message.

ActiveMQ reference guide at  

Jian.Luo wrote:
> Hi All,
> I am wring this email because I am doing a in-house performance tests 
> for ActiveMQ, version 5.3.1, and hit a performance issue.
> The purpose of the test is to measure the scalability of AMQ by
>  -. setting up N pairs of PUB/SUB with N topics. 
>  -. non-durable subscribers
>  -. non-persistent messaging
>  -. Both publishers and consumers are running on Machine A(8 Core/CPU, 
> 32GBRAM, 10GBPS Network);
>  -. Subscribers are runing on Machine B(8 Core/CPU, 32GBRAM, 10GBPS 
> Network).
>  -. ByteMessage is used, each message size is 32k;
>  -. JDK 1.6.21
>  -. Redhat Linux 2.6.18-164.e15 #1 SMP x86_64
> The test results are somewhat surprised me. The best throughput is 
> achieved by
> using only 1 pair of PUB/SUB. And the more pairs of PUB/SUB, the slower
> throughput is.
> I wonder, if any people in this mail list had done similar research around 
> this area. If you have any idea or solution, could you shield me some 
> light
> on it?
> Thank you very much.
>                 ---Jerry

View this message in context:
Sent from the ActiveMQ - User mailing list archive at

View raw message