Lightning Talk: BGP Routing in Kubernetes
summary
Proposes using a custom BGP protocol and a central route feeder to push route changes into Linux iptables instead of having peers delete thousands of entries themselves, shifting routing logic out of per-node churn. Claims this cuts control-plane overhead when scaling from tens to thousands of nodes, and describes a prototype that batches updates (updates after X minutes) to reduce update storms. Prototype is being deployed but not yet presented as fully validated in production.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 0:13BGP
- 0:38Orchestration
- 1:00Ideal Network
- 1:26Two Autonomous Systems
- 1:50Scaling Out
- 2:19What BGP does
- 3:08BGP Deletion
- 3:58Route Feeder
- 4:53Outro
related talks
09mLightning Talk: Hacking Kubernetes Networking to Create Telco Solutions
51mHow Kubernetes Networking Really Works: A Packet’s Journey Across Pods and Nodes
10mLightning Talk: Kubernetes Gateway API for Complex Environments and Service Providers
11mLightning Talk: Intelligent Traffic Routing for Distributed LLM Inference: Beyond Trad...
22mMaking On-Prem Bare-Metal Kubernetes Network Stack Telco Ready
35mLiberating Kubernetes From Kube-proxy and Iptables
10mLightning Talk: Make Way for the Gateway
37mPanel Discussion: Cloud Native Networking State of the Union