CASPaxos has nice latency and recovery times on node failures and partition re-joins. The author of the algorithm has a comparison with other popular consistent

CASPaxos has nice latency and recovery times on node failures and partition re-joins. The author of the algorithm has a comparison with other popular consistent distributed databases https://github.com/rystsov/perseus https://github.com/rystsov/perseus I particularly like novelty of using a function closures which makes it elegant and approachable. I also found the idea of the hierarchical ballots from the article pretty interesting. ([config ballot, range ballot, file ballot, register ballot]). And I suppose in the spirit of the original paper, it would be a nested set of closures :-) One thing that might not be obvious though, compared to other databases, is that reads essentially follow the same path as the writes and cost about the same, and even might result in spreading the update to other nodes that don't have the latest value.