Paxos in Pictures: Lamport's Distributed Consensus Algorithm

Paxos is a computer algorithm to help a network of computers agree on a proposed value. It is hard to understand at first. This video attempts to explain Paxos intuitively yet rigorously. The content comes from Lamport's 2001 paper "Paxos Made Simple".

Note that this video is not as rigorous as the paper it's based on. Specifically, it doesn't contain a precise statement of what Paxos accomplishes and it doesn't say how to optimize Paxos or deal with corner cases. For that, please see the referenced paper "Paxos Made Simple". The intent of THIS video is simply to provide a nice explanation and a visual illustration of the basic idea behind the algorithm.

Slides:

References:

Lamport, Leslie (2001). Paxos Made Simple ACM SIGACT News (Distributed Computing Column) 32, 4 (Whole Number 121, December 2001) 51-58.

(c) 2017 Justin P. Pearson. All rights reserved.

[hangsu]

After searching and learning a number of tutorials about Paxos algorithm, I would have to say this one is the best one for both explanation, example and slides, especially for the beginners. Clean, Simple and easy to understand the general idea of Turing Award winnner's contribution.

[MCDdd]

One of the most clean explanation i've ever seen, not only for Paxos but for computer science topics in general. Thank you.

[harjos]

Best explaination on Paxos ... Amazing clarity... Thank you

[wolfsinem]

by far the best video about the paxos algorithm damn

[rajatpai]

Thank you for doing this.
It is essentially “Paxos made simple, made simple”😃

[casimirrex]

really awesome explanation.. I have gone through many video's and documents. Finally I got an idea.Thanks

[lencumbow]

Best explanation of Paxos I've seen. Thank you.

[Harombidlo]

I love how he almost jucked himself!
Great explanation!

[maryame]

Thank you ! It was really really helpful

[enaxx]

Bestest explanation for paxos ever.

Can we get the slides?

[GameArenaGA]

hi, the explanation is super cool! and could you please upload the subtitle as well? since the auto-generated subtitle is recognized as korean, very confusing and weird. that will greatly help those people who has a little difficulty in catching with your speaking speed or who is non english native speaker, thanks a lot!

[aditya]

When is the proposal saved in acceptors ? after it has learnt that majority has accepted the value rightt?

[Nelo]

Great video but there's one thing I'm not getting.
It seems like once everybody agrees on "carl 4ever" there can never be anything else. While "carl 4ever" can't be undone, there can be another consensus on a 2nd value, right? If this were for a distributed database, you would want the system to get consensus on multiple statements in an order and not have the system only accept one statement ever.
So once everybody agreed on "carl 4ever" how and when does the new round start?

[ManuAnand]

This was pretty awesome. Some minor comments though
1. Style: Loved the "You're Here" icons
2. Style: Maybe you can go a bit easy on "By the way"
3. Content: You mentioned that all nodes start with local unique IDs. I'm not clear about how they are mapped and ordered globally. More concretely, why is Carl's request #2 and Alice's request #1? Which component is doing lexicographical comparison?
4. Content: How long does the system wait to determine that there are no proposals bouncing on the network?

Thanks again for such a great demo!