One of the most important features of ActiveMQ is the ability to set up a network of brokers, and the ability to forward the message to another broker in the network, where there is a demand for messages from a consumer. These brokers could span over WANs and geographies. The messages will get forwarded to the broker where the demand is, which is what makes this a distributed messaging system.
The 'Shared nothing' configuration, where each broker has its own DB instance, is very important. It ensures that every message is accounted for and persisted in the DB to be replayed in case of failure.
Load balancing is important when huge numbers of messages are coming in. The messages get distributed to all the brokers, which are connected. In case of failure of any one broker, the message automatically gets routed to other brokers, ensuring no loss of messages.
By default, the failover protocol uses a random algorithm to choose one of the underlying connectors. If the connection fails, the transport will pick another URI and try to make a connection. The network automatically passes messages to connected brokers that have interested consumers. The failover protocol ensures clients do not need to be manually restarted in the case of a broker failure. As soon as the broker becomes available again, the client will automatically reconnect.
We also appreciate the easy setup of persistent messages using a DB like Oracle.