...
The approach is to create a message broker and from each cluster, one node (coordinator) is registered to the broker. When one node gets an update, the broker is notified and the other clusters are notified by the broker to update their cache. First, you need to install the global cache invalidation feature. For information on how to install a feature, see Installing Features.
You can write a custom notification implementation to support different message broker protocols. The default implementation is the JMS protocol.
...