| <div class="wiki-content maincontent"> |
| <p>We could support a reliable multicast protocol and use that to distribute messages across. This page braindumps how that could fit in with ActiveMQ 4.x's architecture.</p> |
| |
| <ul><li>each VM would have its own embedded broker; JMS connections would talk VM protocol to the embedded broker (or TCP to some remote broker)</li><li>the brokers communicate with each other via multicast |
| <ul><li>the exception to this could be queue dispatch/ack messages could go over UDP rather than mcast.</li></ul> |
| </li></ul> |
| |
| |
| <p>So the brokers communicate with a kind of Bridge which uses the multicast/UDP transports underneath.</p> |
| |
| <h3>Broker keep-alive</h3> |
| |
| <p>Each broker will repeatedly send its own BrokerInfo to the cloud. The MulticastBridge will receive them and keep that broker alive; brokers which don't send the brokerInfo within the heartbeat window are logically considered to have closed.</p> |
| |
| </div> |
| |
