Raft - The Understandable Distributed Protocol

Being an undergrad with no prior distributed systems education, I was able to comfortably read the white paper on Raft. The same cannot be said of any of the highly recommended white papers on Paxos. It's pretty awesome how many implementations have popped up since the white paper, too. I think the cool thing about Raft is that it is making consensus widely available and easy for people to plug into their applications, rather than requiring an expert to integrate Paxos into their application.

There was a presentation about Raft at this years Rubyconf (No ruby involved in the presentation) which may be of use http://confreaks.com/videos/2881-rubyconf2013-raft-consensus...

Note that etcd (part of CoreOS) uses Raft for its distributed log. [1]

[1]: https://github.com/coreos/etcd

Phew! I always thought I was the only one who thought Paxos was incomprehensible. I also recently learned that Google's version on Paxos is much simplified and similar to Raft (mentioned as an aside in talk about Spanner as same conference[1].

[1] http://www.infoq.com/presentations/spanner-distributed-googl...

I read their paper about 5 months ago - seems like a good idea . Specifically, the redundancy portion of the consensus algorithm seemed like a good idea at the time.

Does anyone know if it's implemented anywhere? Or real-life performance statistics and it's overhead cost?

No registration required link to the slides: https://speakerd.s3.amazonaws.com/presentations/7556a1b003d8...

Couldn't help but notice the mixture of genders in the log messages after the whole libuv hubbub over the weekend. Pretty straightforward slides.