Cheap Routes, Fast Paths & Keeping Track
summary
Kubernetes networking techniques combine to cut cross‑AZ egress and shorten hops: use pod topologySpreadConstraints and pod affinity to colocate latency‑sensitive services, enable topology‑aware routing via EndpointSlice hints to bias AZ locality, and apply Istio destination rules plus Kubernetes service internalTrafficPolicy=Local to steer and localize traffic. Demo uses GraphQL → Orders(v1,v2) → Products(v1,v2) across AF South zones, shows 50/50 baseline, then 80% directed to AF South 1A and node‑local product responses (Envoy hoodie), proving the approach works. Tools: EndpointSlice controller, Cilium Hubble, kube‑shark, Jaeger/Prometheus.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 2:41Session overview
- 3:45Kubernetes networking
- 5:08Cost vs latency
- 6:56Network traffic approaches
- 7:17Topology spread constraints
- 9:25Service internal traffic policy
- 11:12Limitations of first approach
- 11:53Topology aware routing
- 13:18ISO destination rules
- 15:23Considerations for ISO rules
- 17:30Demo setup
- 23:28Demo objectives
- 24:29Demo part one
- 27:34Demo part two
- 28:38Demo part three
- 30:30Summary
- 30:51Q&A
related talks
42mKubernetes Networking at Scale
47mFast and Furious Multi-Cloud Connectivity
32mThe Easy--Don't Drive Yourself Crazy--Way to Kubernetes Networking
37mStretching the Mesh
31mTopology-Aware Service Routing for Kubernetes is Coming Soon!
15mSimplify Kubernetes Connectivity
34mLow Latency Multi-cluster Kubernetes Networking in AWS
42mFollow Those Packets!!! Dive Into Network Foundations For Everyone!