Do not consider problem of work load on a single statemachine. we can shard work by having 20 machines each statemachine is assigned to a random group of 5 and then that is the consensus group.
really we want low latency. Node can do the pre-election stuff if it was the last winner.
Having a leader is not really the worst thing becuase should end on hot node
This seamed a reasonably good into to single value paxos https://www.youtube.com/watch?v=d7nAGI_NZPk
Do not consider problem of work load on a single statemachine. we can shard work by having 20 machines each statemachine is assigned to a random group of 5 and then that is the consensus group. really we want low latency. Node can do the pre-election stuff if it was the last winner. Having a leader is not really the worst thing becuase should end on hot node
Some other algorithms to check here https://www.youtube.com/watch?v=XUQJvMALfUA