A few more Paxos papers to take a gander at -- Quorum reads, Compartmentalization, Matchmaking and Pigs. The Paxos family of consensus protocols (and their papers) are as colorful as they ever were.
Has anyone ever written anything that serves just a single file? Should this even be written? I don't know the answer but I've written it, twice.
I run a small experiment to figure out what the optimal setting is for pm2.
A survey of Paxos-based consensus algorithms