Episode

GKE 10 years and SIG Networking, With Antonio Ojea

Podcast: Kubernetes Podcast from Google
Published: Oct 1, 2025
Duration seconds: 2212
Processing state: processed
Canonical source: https://e780d51f-f115-44a6-8252-aed9216bb521.libsyn.com/gke-10-years-and-sig-networking-with-antonio-ojea
Audio: https://traffic.libsyn.com/secure/e780d51f-f115-44a6-8252-aed9216bb521/KPOD261.mp3?dest-id=3486674
JSON: /v1/public/podcasts/kubernetes-podcast-from-google/episodes/gke-10-years-and-sig-networking-with-antonio-ojea
Markdown: /podcast/kubernetes-podcast-from-google/gke-10-years-and-sig-networking-with-antonio-ojea.md

Actions

POST https://stenobird.com/v1/public/podcasts/kubernetes-podcast-from-google/episodes/gke-10-years-and-sig-networking-with-antonio-ojea/transcription-requests
Idempotently request low-priority transcript generation for this episode.
GET https://stenobird.com/podcast/kubernetes-podcast-from-google/gke-10-years-and-sig-networking-with-antonio-ojea.md
Read the agent-friendly Markdown representation of this episode resource.

Summary

Kubernetes core maintainer Antonio Ojea discusses the evolution of networking from hardware-centric models to software-defined APIs. The conversation explores solving IP exhaustion through Multi-Service CIDR and the future of connectivity in multi-cloud environments.

Topics

Kubernetes Networking
CNCF
Gateway API
Multi-Service CIDR
Dynamic Resource Allocation
Cloud Native Infrastructure
Service Mesh
Network Security

Highlights

Main idea: Kubernetes networking is shifting from traditional hardware-based routing to highly abstracted, API-driven models
Technical breakthrough: Multi-Service CIDR addresses the scalability limits of the API server's bitmap by allowing dynamic IP allocation
Failure mode: Over-reliance on large, static CIDR blocks leads to inefficient capacity planning and 'pessimistic' cluster sizing
Practical takeaway: Implementing Dynamic Resource Allocation (DRA) can help manage specialized hardware resources like GPUs and IPs more efficiently
Industry trend: The rise of unauthenticated MCP servers highlights a recurring industry struggle with applying established identity and access management patterns to new scales

Chapters

1:00 CNCF Event Roadmap: An overview of upcoming Kubernetes Community Days (KCD) scheduled for the first half of 2026.
3:30 The Evolution of Networking in K8s: Discussion on why traditional networking approaches were rejected in the core in favor of more abstracted, API-centric models.
6:10 Gateway API and Traffic Management: Exploring how the Gateway API addresses complex traffic routing and network policy challenges at the edge and within the cluster.
9:05 Solving IP Exhaustion with Multi-Service CIDR: A deep dive into the technical challenges of the API server's bitmap and how dynamic CIDR prevents cluster-wide scaling bottlenecks.
12:15 Dynamic Resource Allocation (DRA): How the scheduler can dynamically assign resources like IPs to pods to optimize cluster utilization.
14:55 Testing Networking with Kind: Implementing Gateway API within the 'Kubernetes in Docker' environment for local development and testing.
20:20 Multi-Cloud Connectivity Strategies: Comparing overlay networks versus routing-based approaches to achieve a flat network across disparate cloud environments.
28:25 The Future of Low-Level Networking: Discussing the integration of specialized hardware and the role of technologies like RDMA in high-performance networking.