Keynote: Hello eBPF! Goodbye Sidecars?
Talkat KubeDay Japan 2022 · on Wednesday, 2022-12-07· at 09:30· Level 3 | InterContinental Ballroom·
summary
eBPF loads small programs into the kernel on each node so one instrumented host observes and enforces behavior for every pod without modifying pod YAML or restarting workloads. tools named include Pixie for cluster metrics and Hubble and psyllium for L3–L7 visibility and networking; psyllium implements a sidecarless service mesh with a shared external proxy that cuts packet hops and reduces latency. features cited: ebpf Maps for fast shared state, transparent encryption via IPSec/WireGuard, and limits: kernel programming complexity and managed clusters that block eBPF.
AI-written summary · about
chapters (13)
- 0:00Intro
- 1:03Sidecar model evolution
- 2:39Run custom code in the kernel
- 6:16eBPF doesn't require any app changes
- 7:04Reduce resource usage
- 8:40eBPF observability tools - Cilium Hubble
- 9:09eBPF networking & security
- 10:12Observable, secure, resilient cloud native connectivity
- 10:42Removing sidecars from Service Mesh
- 11:32The network cost of sidecar proxies
- 14:06Kernel delegating complexity to user space
- 15:36Transparent encryption
- 18:03Evolution driven by improved performance
related talks
32mExploring the Power of eBPF for System Observability
23mAn Intro to eBPF with Go : The Foundation of Modern Kubernetes Networking
30mObservability in Kubernetes: Unlocking the Power of eBPF
22mRole of eBPF in Kubernetes Networking and Security
29meBPF on the Rise - Getting Started
29mKubernetes Observability with eBPF
25mBeyond the Buzzword: BPF’s Unexpected Role in Kubernetes
27mStep By Step Kubernetes Observability with eBPF