Meta is using the Linux scheduler designed for Valve's Steam Deck on its servers

Original Article ↗ Hacker News Discussion ↗

Meta is experimenting with a Linux CPU scheduler originally developed for Valve’s Steam Deck, showing that latency-focused optimizations for handheld gaming can also improve efficiency on hyperscale servers. Commenters highlight this as a strong example of open source “cross-pollination,” where work funded for gaming (via Valve and its contractors like Igalia) flows into data centers, just as earlier Meta technologies like the Kyber I/O scheduler have flowed into desktops. The thread also branches into broader points about Linux’s strengths and weaknesses on the desktop, the economics of kernel tuning at scale, and how licensing and ecosystem fragmentation shape who benefits from these advances.

Open source cross‑pollination

  • Thread highlights how Valve’s Steam Deck work (SCX‑LAVD scheduler) is now improving Meta’s server efficiency, and notes the reverse flow (e.g., Meta’s Kyber I/O scheduler helping desktop/SteamOS microstutter).
  • Many see this as “commons” behavior: once code is upstreamed under GPL, it’s no longer “Valve’s thing” or “Meta’s thing.”
  • Some warn against relying on “trickledown” from big firms; corporate priorities can change despite licenses.

Why a handheld scheduler works in hyperscale

  • Commenters are surprised a scheduler tuned for handheld gaming also works for Meta’s servers.
  • Explanation: both gaming and large services have hard latency deadlines (frame times, controller input, voice/video, ad auctions, WhatsApp messaging, etc.), while background work can be delayed.
  • SCX‑LAVD is a latency‑aware scheduler; latency vs throughput is a spectrum, not a simple upgrade path.

Linux scheduling and sched_ext

  • Discussion contrasts the legacy Completely Fair Scheduler (CFS), newer EEVDF, and SCX‑LAVD: each chooses different trade‑offs between fairness, throughput, and latency; none is a universally “strict upgrade.”
  • Linux defaults historically favor throughput/fairness and are hard to tune; at hyperscale, even 0.1% gains justify dedicated kernel engineers.
  • sched_ext (developed at Meta) and BPF‑style mechanisms make it easier to plug in alternative schedulers; SCX implementations live in a shared GitHub repo used by multiple companies.

Valve’s role and contractor model

  • Valve is portrayed as a relatively small, revenue‑dense company that contracts out deep systems work (e.g., Igalia for schedulers, graphics stack, Proton pieces).
  • Igalia is described as a worker‑owned, highly skilled Linux consultancy, seen as a positive example of “company funds OSS” in practice.
  • Several comments argue contracting can work extremely well when scope is tight, expertise is high, and the client remains technically engaged.

Linux ecosystem strengths and weaknesses

  • Many credit Valve (plus earlier Wine/CodeWeavers work) with pushing Linux forward: Proton, DXVK, HDR/VRR on Wayland, Gamescope tools, shader pre‑caching, futex improvements, bcachefs sponsorship.
  • Others stress this builds on decades of volunteer groundwork (Wine, kernel, desktop).
  • Recurrent pain points: desktop Linux UX, accessibility, laptop sleep/hibernate, OOM behavior, hardware/driver quirks, fragmented ABIs and mobile platforms.

Business and ethics angles

  • Meta is criticized for scammy ads and AI misuse but also noted as a major Linux kernel contributor.
  • Valve is praised for technical contributions yet criticized for lootboxes and enabling third‑party gambling around in‑game items; some defend Valve as “least bad,” others call that willful blindness.
  • Side debate on RHEL source availability and GPL obligations, with claims that CentOS Stream effectively exposes the code even if RHEL’s own source distribution is awkward.