After all, right now the perceived norm is to send all your 100Mb/day of data through a kafkaesk array of message queues, stream processors and map-reduce jobs.