[PERF] gmbal objects consuming large part of heap

The gmbal API call in ORB

mom = ManagedObjectManagerFactory.createFederated(MONITORING_SERVER);

is causing a large number of gmbal instances to be created and is affecting GlassFish performance. See [http://java.net/jira/browse/GLASSFISH-17044](http://java.net/jira/browse/GLASSFISH-17044).
Attached is the call stack.

This fix should be to defer the gmbal API calls until there is a JMX client connection.

From Naman: JMX service would start on 8686 (on DAS). So you can check localhost:8686 for the same. From quicklook workspace you can find the sample for the same.

In GlassFish monitoring (StatsProviderManagerDelegateImpl), we defer the gmbal API calls by extending MBeanListener.CallbackImpl and overriding mbeanRegistered method. This method is called when AMX DomainRoot is loaded (when there is a JMX, for example). AMX DomainRoot needs to be ready before any other mbeans can be registered.

Fix should be targeted for GlassFish 3.1.2.