I'm trying to put together a Samza subsystem that is able to auto-size a Samza job by looking at metrics. Here are some initial thoughts: https://docs.google.com/document/d/1gqGuc2PGBvjbpE2wP-DbOTujSuDQjPwpYN9guWZde00/edit?usp=sharing
It'd be great to get your feedback. -- thanks rayman