Skip to content

QEMU in the Fast Lane: Accelerating KubeVirt Networking With eBPF#

KubeCon + CloudNativeCon North America 2025 digest Program Core pick 10–13 November 2025

Recording

Daniel Borkmann & Anton Protopopov, Isovalent at Cisco — Core

Play the recording: QEMU in the Fast Lane: Accelerating KubeVirt Networking With eBPF

  • Why: VM-on-Kubernetes convergence keeps colliding with the latency/throughput tax of the virt networking path. The usual escape hatch, SR-IOV, costs you host-side observability and policy enforcement. This is a contributor-led internals talk from the people who build the Cilium datapath.
  • What: An architecture that puts QEMU's network path on AF_XDP to get near-line-rate KubeVirt networking while retaining host-level visibility and policy. (from abstract only)
  • How: eBPF/AF_XDP fast path built on Cilium, netkit, and upstream Linux kernel enhancements, with contributions landed in both QEMU and the kernel to back KubeVirt pods with high-performance AF_XDP interfaces. (from abstract only)
  • Where applicable: KubeVirt/QEMU-KVM on Kubernetes with a recent kernel and Cilium + netkit; general technique, but depends on the upstream kernel/QEMU work being present.