The pubsub_clustering branch in SVN has been updated with the changes required to enable the pubsub module in clustering mode. The new clustering capabilities include the following:
1) A local in-memory cache of newly published items (write cache)
2) A replicated cache for recently published items (read cache)
3) Several new cluster tasks for maintaining pubsub state across members
In addition to the changes in the pubsub module, we have made the following changes to improve the clustering capabilities of Openfire:
1) The clustering plugin has been updated to support the latest Coherence release (3.7.1)
2) A new plugin (hazelcast) has been created to enable Openfire clustering using Hazelcast Community Edition (CE)
Hazelcast is an open source (Apache 2.0) clustering and data distribution framework. The new plugin is designed to be a drop-in replacement for the existing clustering plugin, providing an equivalent level of service for a large-scale Openfire deployment.
These changes are ready for testing now. Note that the pubsub_clustering branch must be merged back into the SVN trunk to ensure that the pubsub clustering enhancements are included in subsequent OF release(s).