Could the performance numbers be possibly related to application level code? There are a couple of JMS load test tools available, I like to use the maven-activemq-performance-plugin http://activemq.apache.org/activemq-performance-module-users-manual.html
as it supports testing for many different use-cases, including batched transactions. On Oct 21, 2011, at 3:49 AM, lzr wrote: > I've tested consume the messages without the jms.useCompression in the > connection: > Producer took 55 seconds (5 seconds to send them with > jms.useCompression=true) to send all the 10000 messages. > Consumer used 80 seconds to receive all the 10000messages. > The consumer performance is same as before. It seemed that this parameter > didn't affect the message receiving. > > > > > > At 2011-10-21 08:38:50,lzr <jsw...@163.com> wrote: >> Have you ever test the performance of the "consumer.receive"? >> >> >> >> >> >> At 2011-10-20 17:23:30,SuoNayi <suonayi2...@163.com> wrote: >>> I have tested the performance of the consumer before some days, it 's not >>> possible consumer is slower than producer so much. >>> you can try to set messagelistener for consumer instead of using the >>> synchronous method consumer.receive(). >>> >>> At 2011-10-20 16:49:18,lzr <jsw...@163.com> wrote: >>>> Yes, I run the both consumer and producer in Eclipse. When the sent done, >>>> start consumer to receive the messages from queue. >>>> >>>> >>>> >>>> >>>> >>>> At 2011-10-20 16:20:09,lzr <jsw...@163.com> wrote: >>>>> Yes, its are persistent messages. >>>>> As these messages in a huge transaction, without compression property set >>>>> the broker will run out of memory. >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> At 2011-10-20 15:42:28,"Torsten Mielke" <tors...@fusesource.com> wrote: >>>>>> Are these persistent messages? >>>>>> >>>>>> From my own performance tests I know that enqueuing a persistent message >>>>>> takes longer than dequeuing it, which is the opposite behavior from what >>>>>> you're seeing. Persistent msgs need to be persisted to the brokers >>>>>> store. >>>>>> >>>>>> Do you encounter the same behavior when turning off compression? >>>>>> Are producer/broker/consumer on the same machine? >>>>>> Also, I presume there are no transactions used? >>>>>> >>>>>> >>>>>> >>>>>> Torsten Mielke >>>>>> tors...@fusesource.com >>>>>> tmie...@blogspot.com >>>>>> >>>>>> On Oct 20, 2011, at 9:32 AM, lzr wrote: >>>>>> >>>>>>> Hi all, >>>>>>> >>>>>>> >>>>>>> In my cases, I produces and consumes messages with big data: 40K per >>>>>>> message, 10000 messages produced and consumed for each exchange. For >>>>>>> better performance I used jms.useCompression in my connection and it >>>>>>> did work well: >>>>>>> 10000 messages sent in 5 seconds; >>>>>>> But my consumer takes about 70 seconds to receive these messages from >>>>>>> queue (Just received, no process to message). >>>>>>> How is the consumer so lower than producer? Is there any optimization >>>>>>> to the consumer? >>>>>>> >>>>>>> >>>>>>> It's much appreciated if any suggestions! >>>>>>> >>>>>>> >>>>>>> Best regards, >>>>>>> Zhuran Li >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> Torsten Mielke tors...@fusesource.com tmie...@blogspot.com
