As you mention, you’ll have a pool of MDBs that will be publishing to the Broker. You effectively already have a connection pool. As long as you have logic in the bean to connect to the Broker (either at start-up or on first activity) and to reconnect automatically if the connection is dropped, using your pool of MDBs to control how many connections to the Broker you have should work just fine. Each bean should use the same connection parameters when connecting the Broker.