Data processing with Creation-Expensive Beans

[deutinger69]

Hello,

I need a design hint for the following application scenario:

I have to process data in near real time which arrives in short intevalls from various identical sources. Before the data can be processed by an object (Session or POJO) the object has to set-up its internal configuration data. The setup procedure for this configuration data is more epensive than the final data processing.

What would be the best pattern to get this done?

1) Simply using stateful beans seems to be no option because there are too many external data sources and the external data reciever must not wait for the data processing to finish.

2) Ideal would be MDBs, but how can I make sure that they are properly reused and there are not too many new instances created, as this is quite expensive and memory consuming?

Is there anyone with a bright idea or someone who has solved a similar situation?

Best regards
Alex

[camel]

You can limit the size of your MDB pool through JBoss's configuration. It probably ships with a decent, conservative default setting so you won't exhaust your resources. I don't remember the details of changing this, but it's in the free documentation: http://www.jboss.org/docs/index, and probably the wiki as well.

JBoss should automatically handle reusing the MDB instances.

[deutinger69]

@camel
This is the solution I consider at the moment. I'm just wondering whether someone has *the* brilliant idea - cause class variables or singletons are no option for MDBs.

[genman]

There is the famous SingletonMessageDrivenBean -- take a look at standardjboss.xml in JBoss 3.2.5. Or, store your crap in a MBean, which is effectively a singleton.