Kubernetes Podcast from Google: KubeCon EU 2025 - Comprehensive Summary

Release Date: May 15, 2025
Hosts: Abdel Sghiouar, Kaslin Fields
Guests: Various Kubernetes community members and experts

Introduction

In the KubeCon EU 2025 episode of the Kubernetes Podcast from Google, hosts Abdel Sghiouar and Kaslin Fields delve into the latest happenings and innovations within the Kubernetes community. The episode encapsulates live-streamed interviews from the conference floor in London, offering listeners a curated selection of conversations that cover platform engineering, cutting-edge technologies, core Kubernetes updates, and unique user stories.

News Highlights

1. Automated Governance Maturity Model by CNCF

• Host Summary: The Cloud Native Computing Foundation (CNCF) introduced the Automated Governance Maturity Model developed by CNCF's Technical Advisory Group (TAG) Security.
• Purpose: Assists organizations in evaluating and enhancing governance policies with a focus on automation amidst rapid development and increased AI system usage.
• Key Points:
  • Ensures systems adhere to organizational expectations.
  • Facilitates compliance with regulations.
  • Embeds automation into traditional governance tasks.

2. Kubernetes 1.33 Release Features

• Host Summary: The release of Kubernetes 1.33 continues to generate buzz with new features such as:
  • Dynamic Resource Allocation (DRA)
  • Image Volumes
  • Horizontal Pod Autoscaling
• Recommendation: Interested users are encouraged to explore the Kubernetes IO blogs for detailed feature insights.

3. Open Observability Summit

• Host Summary: Formerly known as Open ObservabilityCon, the event has been renamed to Open Observability Summit to prevent confusion with similarly named events.
• Details:
  • Date & Location: June 26, 2025, in Denver, Colorado.
  • Objective: Unites developers, operators, and business leaders to enhance open-source observability projects and practices.

Platform Engineering

1. NAV's NICE Platform with Hans and Adun

• Guests: Hans and Adun from NAV, the Norwegian Labor and Welfare Administration.
• Discussion Highlights:
  • NICE Platform Overview:
    • A Kubernetes-based application platform initially deployed on-premises and later transitioned to Google Kubernetes Engine (GKE).
    • Acts as an abstraction layer with custom Kubernetes operators and manifests.
  • Community Building:
    • Established a robust platform engineering community in Norway, comprising thousands of members across 60-70 companies.
  • OpenTelemetry Integration:
    • Implemented auto-instrumentation of OpenTelemetry to standardize observability across diverse applications.
    • Quote:
      Hans [10:30]: "This allows us to instrument once and run anywhere, eliminating the constant need to switch libraries and agents."

2. Insights from Andy and Max on Platform Engineering for Architects

• Guests: Andy and Max, authors of Platform Engineering for Architects.
• Discussion Highlights:
  • Philosophy of Platform Engineering:
    • Treating platforms as products by identifying and solving developer pain points.
    • Emphasizing user-centric design to ensure platform adoption.
  • Evolution from DevOps:
    • Moving beyond traditional DevOps practices to create more integrated and scalable platforms.
  • Key Takeaway:
    • Max: "Start by understanding the pain points and build a self-service platform that addresses those needs seamlessly."

LinkedIn's Scalable Compute Platform

Guests: Amit and Ronak from LinkedIn

• Discussion Highlights:
  • Scalable Compute Platform:
    • Managing bare metal servers and deploying applications via Kubernetes.
    • Leveraging Kubernetes Operators and Custom Resource Definitions (CRDs) for automation.
  • Node Lifecycle Management:
    • Critical for handling demanding AI and machine learning workloads.
    • Quote:
      Ronak [20:45]: "Correctness and scalability of operators are paramount given our massive scale requirements."
  • Multi-Cluster Orchestrator (MCO):
    • Announcement of an open-source tool for managing multi-cluster environments.
    • Amit: "We're excited about MCO and look forward to community feedback as it evolves."

AI and Machine Learning on Kubernetes

1. Running Large Language Models (LLMs) on Kubernetes

• Hosts: Abdel and Mofie
• Discussion Highlights:
  • Why Run LLMs on Kubernetes?
    • Use Cases:
      • Organizations requiring data residency and control.
      • Cost-effective scaling by pre-provisioning resources.
    • Technical Insights:
      • Auto-scaling based on custom metrics like tokens per second rather than just CPU or request counts.
      • Quote:
        Abdel [35:50]: "Scaling based on tokens or GPU usage ensures optimal resource allocation for AI workloads."

2. Gateway API Inference Extension

• Discussion Highlights:
  • Features:
    • Multi-modality routing based on user requests (e.g., text summarization, image processing).
    • Model-based routing to direct requests to specific backends.
    • Intelligent routing decisions using custom metrics from inference servers.
  • Future Prospects: Early access and continued development with community contributions.
  • Quote:
    Abdel [38:15]: "The Gateway API inference extension is set to revolutionize how we interact with AI models by enabling smarter routing mechanisms."

Kubernetes Project Updates

1. SIG ETCD with Ivan Valdez

• Guest: Ivan Valdez, Co-Chair of SIG ETCD
• Discussion Highlights:
  • ETCD 3.6 Release:
    • First minor release in four years, addressing maintenance and feature updates.
  • ETCD Operator:
    • Introduction of an official ETCD Operator to simplify running standalone ETCD clusters within Kubernetes.
    • Quote:
      Ivan [50:10]: "The ETCD Operator is in its alpha stage, and we're actively seeking contributors to enhance its capabilities."

2. Insights from Dzego McLeod on Kubernetes Evolution

• Guest: Dzego McLeod, Engineering Director at Google
• Discussion Highlights:
  • Kubernetes and AI/ML Workloads:
    • Enhanced awareness of hardware accelerators.
    • Collaboration with frameworks like Slurm, RAY, and Run AI.
    • Quote:
      Dzego [55:30]: "Kubernetes is evolving to become more hardware-aware and is integrating seamlessly with specialized AI/ML frameworks."
  • Future Directions:
    • Development of AI agents to simplify Kubernetes interactions, reducing the need for direct kubectl usage.
    • Dzego: "We anticipate AI agents playing a pivotal role in abstracting Kubernetes complexities for end-users."

Unique User Stories

Clement's Dual Role: Banking and Farming Automation

• Guest: Clement, Software Engineer at Post Finance & Farmer
• Discussion Highlights:
  • Banking Role:
    • Managing the Kubernetes platform for Post Finance, including provisioning clusters and migrating from kubeadm to Cluster API.
    • Quote:
      Clement [1:05:20]: "Migrating to Cluster API allows us to manage our clusters more efficiently without downtime."
  • Farming Automation:
    • Utilizing Kubernetes and Prometheus to automate a milk dispensary and monitor cows.
    • Setup includes a four-node Kubernetes cluster at home running on Dallas Linux, integrated with Prometheus for monitoring.
    • Clement: "Using Kubernetes on the edge has enabled us to maintain consistent milk availability through automated alerts and monitoring."

KubeCon EU 2025 Overview

• Attendance: Over 13,000 attendees, marking it as the largest KubeCon to date.
• Venue Layout: The show floor was divided into two sections to accommodate the large crowd, reducing congestion but increasing walking distances.
• AI Presence: A significant focus on AI, with numerous keynotes and project showcases emphasizing AI and machine learning integrations within Kubernetes.
• Organizational Insights:
  • Hosts' Reflection: The increased frequency of KubeCons (five per year globally) raises concerns about potential organizer burnout but also offers more opportunities for regional engagement.
  • Quote:
    Kaslin [1:15:45]: "With KubeCon happening almost every two months worldwide, we're excited to see diverse innovations from different regions."

Conclusion

The KubeCon EU 2025 episode of the Kubernetes Podcast from Google offers a deep dive into the dynamic Kubernetes ecosystem. From platform engineering advancements and scalable compute solutions at LinkedIn to unique applications like farm automation and the integration of AI/ML workloads, the episode underscores Kubernetes' pivotal role in modern infrastructure. Additionally, updates from key project contributors like SIG ETCD and insights into Kubernetes' evolving support for AI workloads highlight the community's continuous innovation and collaboration. Whether you're a seasoned Kubernetes practitioner or new to the ecosystem, this episode provides valuable perspectives and actionable insights to navigate the ever-expanding Kubernetes landscape.

For more detailed information, including transcripts and show notes, visit kubernetespodcast.com. To stay updated, follow Abdel and Kaslin on Twitter @KubernetesPod or reach out via email at kubernetespodcast@google.com.