Visualization of a Paxos-based distributed state machine

http://rystsov.info/2016/05/01/paxos.html

• “Paxos made simple” paper (2001) - algorithm how to transform any FSM into a fault-tolerant distributed system:
  • Build linearizable distributed write-once register (based on his Sinod protocol)
  • Build linearizable distributed append-only log (sequence of write-once regs)
  • Use distributed append-only log to store the sequence of all events leading a system to the current state instead of storing the state itself
• logs - is natural, but need GC or snapshots. Something else to model FSM?
  • TreodeDB used Synod-based distributed rewritable register
  • riak_ensemble (?)
  • proved FSM over distributed variable
• Log-based solutions are more popular because they had the advantage of time. Until Vertical Paxos (2009), log replication was the only consensus protocol capable of cluster membership change -> the only solution.
• Raft’s joint consensus membership change