Supercharge Kubernetes Networking with Cilium and IPv6
summary
Cilium runs on a kind two-node IPv6-only Kubernetes cluster using Cilium IPAM to assign pod /128s and native routing to program host routes instead of tunneling; kube-proxy is replaced and services get single-stack IPv6 addresses with CoreDNS returning AAAA records. Hubble (eBPF) shows per-connection events and explicitly logs policy-deny drops. Enabling Big TCP raises gso_max from 64k to 192k and boosts performance: min latency ~71 µs, P90 ~225 µs, P99 ~697 µs and ~6,000 req/s observed. native routing Big TCP Hubble observability
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:35Cluster Configuration
- 4:03Native Routing
- 5:50Cilium Demo
- 7:00Sample Workload
- 7:42Test Connectivity
- 11:02Hubble
- 11:51Enable Hubble
- 12:23Observe Hubble traffic
- 15:00Create Network Policy
- 19:13Big TCP
- 23:02Big TCP for IPv6
related talks
26mIPv6 First, Not Just Ready: Kubernetes Without IPv4 Using Cilium
07mWelcome + Opening Remarks
31mUnleashing the Tides of Kubernetes Networking by Removing kube-proxy
35mNext-level Kubernetes Networking with Cilium - Nico Meisenzahl and Philip Welz
40mOur Experiences Deploying Kubernetes with IPv6
25mThat’s Just My Cup of Tea: Configuring Cilium for Performance and Scale
29mSimplifying the Networking and Secur...
24mThe Cilium Story - Why We Created Cilium