Local Routing in a new Indefinitely Scalable Architecture [pdf]

cs.unm.edu

39 points·mr_tyzic··3 comments

I'll always upvote Dave Ackley. His vision for computing is so audacious it seems unbelievable. Global consistency? Don't need it! Determinism? An illusory safety blanket for wimps and suckers! Sure, everyone talks a big game about eventual consistency, but who's signing up to use an eventually consistent CPU? Dave freakin' Ackley, that's who.

And the whole thing would sound like blue-sky bullshit except he's out there actually building it. I think the best place to start is this demo of a network switch made out of cells: https://www.youtube.com/watch?v=I4flQ8XdvJM&t=11m23s (and then watch the rest of the video if you liked that)

His more recent videos are weekly updates on a "T2 Tile" prototype made out of 3d printed parts, custom PCBs, and like a hundred BeagleBones: https://www.youtube.com/c/TheT2TileProject

Also, all his code is on GitHub: https://github.com/DaveAckley/

You might be interested in The Agoric Papers [1]

[1] https://e-drexler.com/d/09/00/AgoricsPapers/agoricpapers.htm...

Why mock eventual consistency though? I get it, some megacorps and VC funded startups are pushing strong consistency and attacking other consistency model, but it's just aggressive PR detached from reality. If you try various consistency models, you'll find the coordination-free camp of consistency models to be much nicer and much easier to work with, reason about and strong consistency is not actually all that strong and more like a lie in every single practical deployment that has to interact with humans and they always have to build eventual consistency on top if they actually need some guarantees.

Reply on news.ycombinator.com