Open sebpiq opened 9 years ago
Maybe redis is not the solution : http://stackoverflow.com/questions/1919472/low-latency-large-scale-message-queuing.
Or maybe we don't need a separate queue, but rather proxying directly the messages from one server to the next. Each server registers itself to the others, so they can communicate together. Or maybe it could be a setting rhizomeInstances
with IPs of all other rhizome servers running ... but in any case, we need a centralized database to save persistent info
Another option might be the cluster module built into node, and each cluster keeping track of its own connections like now.
I don't think this would work. If a client disconnects and reconnects, it might be handled by a different server of the cluster, which doesn't have infos about the connection.
Besides, I want also to be able to run servers on different machines for maximum scalability.
Cool, the cluster thing was just an off the cuff idea.
Now redis persistence is available, which should solve the cluster problem, and allow this to move forward.
Look into using redis for pub/sub instead of the ConnectionManager.
This would allow to scale, and have several rhizome servers communicating together through a queue system.