In consistent hashing, why always go to the server with the *next* hash instead of the *closest* one?
ChatGPT tells me that it's to avoid having to deal with "tie breaking" for equidistant points, but that seems like a trivial obstacle.
Intuitively, closest hash makes more sense. For example, if there are only 2 servers:
- With closest-hash matching, each will get 50% of the requests.
- With next-hash matching, there's a 50% chance that one server will get >=75% of requests.
