This talk was given SRECon EMEA in Dublin.
Traditionally, service discovery has leaned towards strong consistency. If you are querying an endpoint, ideally you don’t have to deal with split brain on the set of active healthy nodes.
I got inspired to give this talk after chatting with a group of colleagues and debating whether strong consistency is ideal if you are operating at the thousands or tens of thousands scale of endpoints and things are constantly moving around.
The talk was geared towards introducing systems like ZooKeeper and Raft and explaining consistency levels. If you need to store state, then a strongly consistent system is ideal. For service discovery though, an eventually consistent approach will most likely scale better.
You can find the slides on GitHub