Skip to content

SYSTEM Cited by 1 source

QEMU

QEMU (Quick EMUlator) is the long-standing open-source machine-emulator-and-virtualizer that, paired with Linux KVM, is the reference conventional-hypervisor stack in the Linux world. Broad device emulation surface, supports PCI passthrough, is the default workhorse behind libvirt / OpenStack / oVirt. Wiki appearance here is a reference / wiki-touchpoint page, not a deep dive.

Seen in (wiki)

  • Fly.io evaluated and rejected QEMU for GPU Machines. "We like QEMU fine, and could have talked ourselves into a security story for it, but the whole point of Fly Machines is that they take milliseconds to start. We could not have offered our desired Developer Experience on the Nvidia happy-path." QEMU is on Nvidia's supported driver path; Fly.io could have used it and gotten working Nvidia drivers out of the box — but at the cost of boot-time orders of magnitude slower than Firecracker-class micro-VMs. Fly chose Cloud Hypervisor + off-path driver integration instead. (Source: sources/2025-02-14-flyio-we-were-wrong-about-gpus)

Why it matters (wiki context)

  • Nvidia "works on QEMU" — Nvidia's driver team engineers for QEMU (and VMware). Any micro-VM-shaped VMM needs to either sit on QEMU, impersonate QEMU (Fly.io hex-edited closed-source Nvidia drivers "to trick them into thinking our hypervisor was QEMU"), or negotiate directly with Nvidia. None of those three paths is turnkey.
  • Boot-time cost. QEMU boots on the order of seconds, not milliseconds. For serverless / ephemeral / per-request compute products, this eliminates QEMU from contention. Fly.io's fast-boot DX is a canonical example.
Last updated · 200 distilled / 1,178 read