3 ms·
I haven't found a problem well suited to either as a solution. Either the performance constraints are too tight for industrial applications or you are in a P2P
by blamestross 7y ago
I haven't found a problem well suited to either as a solution. Either the performance constraints are too tight for industrial applications or you are in a P2P space and they are both predicated on nodes not being hostile so you can't use them.
In practice it just seems most efficient to be tolerant of consensus failures and focus on cAP.
- dilyevsky 7y agoA lot of real use production systems (spanner, cockroach, tidb) use the opposite approach - sacrifice reliability for consistency. Scaling constraints are usually solved via sharding (running multiple raft/paxos fsms per dataset)