You may have tried this out but increasing the maxMessagesPerSession activation config property will increase the prefetch size and generally speed it up a bit more. I assume u have already increased the maxSessions activation config property to increase the no of parallel Mdb instances. Here is a link for the properties that AMQ allows http://activemq.apache.org/activation-spec-properties.html. The configurable OpenEJB properties are given here http://openejb.apache.org/example-generated-documentation.html.
Try setting the InstanceLimit property of the MdbContainer to 0 so that the no of instances created matches the no of AMQ sessions available. For setting this you need to set this as a system property. The property should be containerId.InstanceLimit where containerId is of the format <artifactId>.<Resource Group Name>-<listener interface> eg: org.apache.geronimo.configs/activemq-ra/2.2-SNAPSHOT/car.ActiveMQ RA-javax.jms.MessageListener ie <artifactId> = artifactId of the jms RA <Resource Group Name> - The resource Group name u gave while creating the RA <listener interface> - javax.jms.MessageListener in this case So the property in this case can be set as org.apache.geronimo.configs/activemq-ra/2.2-SNAPSHOT/car.ActiveMQ\ RA-javax.jms.MessageListener.InstanceLimit=0 Regards Manu On Sun, Mar 23, 2008 at 4:49 PM, the666pack <[EMAIL PROTECTED]> wrote: > > hello, > > i am trying to test the performance of geronimo under load for message > driven beans. the observed behaviour is that i get a maximum of about 85 > transactions per second. after this, tested with about 60 concurrent users, > the performance drops again to about 45 tps. i hoped to increase the > performance somehow.. things i already tried: > > -increase the mdb pool size > -increase the thread pool size > -increase the connector-thread-pool-size > > unfortunately nothing helped, the performance situation remains the same. > > another observed behaviour is that in the time of silence between tests the > server needs some time to fulfill the requests sent before. so i see it is > still working although i am not sending mdb-requests anymore. > > i suppose these are the requests that are still in the queue and have to be > finished. i think this is the reason for the performance drop with more > concurrent users. i would like to see it perform without this > "aftertime-working" and handle the requests "in time". > > maybe someone can give me a hint which other values could be important to > change for the performance of mdbs getting real good. i thank you very much > for your help, > > greetings, > > mario > -- > View this message in context: > http://www.nabble.com/MDB-performance-tuning%2C-configuration-tp16234678s134p16234678.html > Sent from the Apache Geronimo - Users mailing list archive at Nabble.com. > >
