Journalled JDBC Master/Slave Performance Issue

[panu <pa...@team.telstra.com>]

Hi Guys,

   I set-up Journalled JDBC Master/Slave using AMQ SNAPSHOT 5 (the latest in
the download list) and Oracle (also MySQL) database. I set up experiment
with one producer, one consumer, one queue, and 100 of 1K messages. The
producer and the consumer communicate with AMQ using Stomp protocol. I run
two cases. First, the producer sends the messages, and then after the
producer is done, the consumer is run to receive the messages. Second, I run
the producer and the consumer at the same time (which is our real-life
scenario). I found that the throughput of the second case is much less (100
messages / ~10 seconds) than that in the first case (100 messages / ~less
than a second). I check the log and notice that Journal checkpoints every
time a message is added and removed from the queue in the second case while,
in the first case, only a few checkpoints (I think two) are made. I also
check the JournalPersistenceAdaptor.java (just a bit), and it seems to me
that doCheckpoint() is always called with fullCheckpoint set to true every
time in the second test case.

   If someone please confirm whether these are the expected behaviour of
Journal? If they are, then I will try to find another way to address the
requirements we have, instead of using Journalled JDBC Master/Slave.

   PS: Snapshot of the log files are attached:
   http://www.nabble.com/file/p12219809/activemq.log.case.01
activemq.log.case.01 
   http://www.nabble.com/file/p12219809/activemq.log.case.02
activemq.log.case.02 

Regards,
Panu
-- 
View this message in context: http://www.nabble.com/Journalled-JDBC-Master-Slave-Performance-Issue-tf4292542s2354.html#a12219809
Sent from the ActiveMQ - User mailing list archive at Nabble.com.