Migrating from DPDK to AF_XDP for High-Performance Networking in...
summary
Claim: existing DPDK apps run on AF_XDP with no application code changes by using the --vdev switch to select the AF_XDP PMD, and orchestrate eBPF lifecycle via BPFman CRDs + CSI plugin so BPF bytecode ships as OCI images, gets signed with cosign, and pins maps into a per‑pod BPFFS. The device plugin + CNI offloads privileged load/unload and removes the UDS handshake, the team migrates an OAI UPF successfully, tested on kernels >5.19, and exports metrics via OpenTelemetry.
AI-written summary · about
chapters (5)
- 0:00Introduction to BPF man
- 2:39Managing eBPF in Kubernetes
- 7:48Introduction to AF_XDP
- 16:32Integrating BPF man and AF_XDP
- 19:42Questions and Answers
related talks
18mOpen vSwitch with AF_XDP as a userspace datapath
25mDPDK performance optimization for edge computing scenario
37mEfficient and Flexible Virtual Machine Networking Through eBPF
28mLow-Latency, Deterministic Networking with Standard Linux using XDP Sockets
36mBringing vDPA to Life in Kubernetes
22mIntegrating Cloud Native Data Plane (CNDP) with OMEC BESS-UPF
22mKernel-Native Packet Processing on AMD GPUs: XDP Offload without Userspace Runtimes
36mRunning High-performance User-space Packet Processing Apps in Kubernetes