Hacker News new | ask | show | jobs
by expede 1131 days ago
An absolute banger of a paper! (Hellerstein is a coauthor, so immediately expected to be good). Definitely some stuff in here that's further out, but a very clear articulation (or roadmap!) of where distributed systems could go.

Verified lifting is very cool but needs lots of research, though I'm unclear that's a hard requirement for this vision to become reality.

1 comments

You're pretty hyped about this, would you mind providing more details about the sections you find interesting? I've yet to read the paper.
Snippets like this do it for me

    𝑐𝑜𝑠𝑡(add_person, 𝑛𝑖) ≤ 0.01. The cost of running add_person
    on 𝑛𝑖
    instances of type 𝑖 machines must be less than the
    specified value. The value can either be specified by the
    end user or provided by the hosting platform.

    𝑖 𝑛𝑖 > 0. Allocate some machines to fulfill the workload.