In my experience, persistent message queue is just a poor secondary database. If anything, I prefer to use ZeroMQ and make sure everything can recover from an

In my experience, persistent message queue is just a poor secondary database. If anything, I prefer to use ZeroMQ and make sure everything can recover from an outage and settle eventually. To ingest large inputs, I would just use short append only files and maybe send them over to the other node over ZeroMQ to get a little bit more reliability, but rarely are such high volume data that critical. There is nothing like free lunch when talking distributed fault tolerant systems and simplicity usually fares rather well.