At first I was also a little confused by this architecture. I think that was because I was missing a clear picture of what they want to do. I think it is: - on

At first I was also a little confused by this architecture. I think that was because I was missing a clear picture of what they want to do. I think it is: - one leader that can write to the db - a ton of other hosts that can read from the "streamed" db and want to do so often/fast - hosts send their writes (if they have any?) to the leader via some api call (?) Is that right? If so, the solution makes a ton of sense. "Just use pg/MySQL" in that case would mean hundreds (thousands?) of read replicas, which doesn't sound fun, or every read over the network.