← Back to context

Comment by kstrauser

21 hours ago

I'm more surprised that the scheduler made for a handheld gaming console is also demonstrably good for Facebook's servers.

9 comments

kstrauser

Reply
giantrobot  20 hours ago

Latency-aware scheduling is important in a lot of domains. Getting video frames or controller input delivered on a deadline is a similar problem to getting voice or video packets delivered on a deadline. Meanwhile housecleaning processes like log rotation can sort of happen whenever.

bigyabai  21 hours ago

I mean, part of it is that Linux's default scheduler is braindead by modern standards: https://en.wikipedia.org/wiki/Completely_Fair_Scheduler

  • 3eb7988a1663  20 hours ago

    Part of that is the assumption that Amazon/Meta/Google all have dedicated engineers who should be doing nothing but tuning performance for 0.0001% efficiency gains. At the scale of millions of servers, those tweaks add up to real dollar savings, and I suspect little of how they run is stock.

    • Anon1096  20 hours ago

      This is really just an example of survivorship bias and the power of Valve's good brand value. Big tech does in fact employ plenty of people working on the kernel to make 0.1% efficiency gains (for the reason you state), it's just not posted on HN. Someone would have found this eventually if not Valve.

      And the people at FB who worked to integrate Valve's work into the backend and test it and measure the gains are the same people who go looking for these kernel perf improvements all day.

  • accelbred  21 hours ago

    CFS was replaced by EEVDF, no?

    • 0x1ch  20 hours ago

      I vaguely remember reading when this occurred. It was very recent no? Last few years for sure.

      > The Linux kernel began transitioning to EEVDF in version 6.6 (as a new option in 2024), moving away from the earlier Completely Fair Scheduler (CFS) in favor of a version of EEVDF proposed by Peter Zijlstra in 2023 [2-4]. More information regarding CFS can be found in CFS Scheduler.

    • jorvi  18 hours ago

      Ultimately, CPU schedulers are about choosing which attributes to weigh more heavily. See this[0] diagram from Github. EEVDF isn't a straight upgrade on CFS. Nor is LAVD over either.

      Just traditionally, Linux schedulers have been rather esoteric to tune and by default they've been optimized for throughput and fairness over everything else. Good for workstations and servers, bad for everyone else.

      [0]https://tinyurl.com/mw6uw9vh

    • phdelightful  21 hours ago

      Parent's article says

      > Starting from version 6.6 of the Linux kernel, [CFS] was replaced by the EEVDF scheduler.[citation needed]