CNCF-hosted Co-located Events Europe 2025

As an old-school DBA, maybe you still don’t like the idea of running mission-critical databases on Kubernetes.

I can understand that – you’ve spent decades learning your trade, honing your database expertise and putting together your database administration toolbox. You’re confident in your ability to implement and maintain a reliable, secure, performant database environment. Why would you risk upsetting that by migrating to a completely different architecture?

Also, everyone tells you that Kubernetes is only for stateless applications. Why on earth would you want to run your databases there?

The database landscape is changing. For several years now, I’ve been working with many customers who run mission-critical, multi-terabyte databases on Kubernetes. Some of them look after many hundreds or even thousands of databases.

The aim of this session is to show you how Kubernetes can enhance your database expertise and give you powerful new tools to solve old challenges.

Is your organization scaling Apache Kafka on Kubernetes with the need for efficient resource management and automated balancing? Strimzi, an open-source solution for running Kafka on Kubernetes, integrates with Cruise Control to offer advanced workload rebalancing, monitoring, and optimization. This integration provides you with production-grade automation for workload distribution and resource efficiency, reducing manual intervention and improving resiliency in a cloud-native environment.
This session is for Kafka operators, SREs, and DevOps engineers seeking to optimize Kafka on Kubernetes. We’ll cover the setup of Cruise Control, automated rebalancing strategies, and real-world examples demonstrating dynamic traffic scaling, rapid failure recovery, and optimized resource usage. Attendees will also see a live demo of Cruise Control rebalancing a Kafka cluster to maximize performance and stability.

Developers of AI applications usually find thmeselves stuck in unexpected errors when migrating testing workload onto Kubernetes. One of the main reasons is that accessing remote storage is complicated and unstable. More specifically, developers are often puzzled by the following questions:
- Why significant extra resource is required, such as memory and disk I/O?
- Why do workloads that seem work well suddenly fail to access remote storage and hard to recover.
- Why can't I add another data source for new experiments without restarting online IDE?
Fluid is a CNCF project focusing on data accessing and workload orchestrating for AI applications. Fluid aims to assist users in building AI platforms based on Kubernetes. In this talk, we'll discuss why the questions are frequently asked based on Fluid users' feedback and how Fluid try to solve them.

GPU-focused workloads such as cloud gaming and AI inference jobs demand access to data at scale, which is often challenged by prohibitive egress costs and latency. This talk introduces a system and strategy designed to accelerate and optimize S3 object storage performance across a global fleet of edge Kubernetes clusters. The solution is built on Kubernetes, leveraging AIStore with OVN, Metallb, and many other common OSS components. Attendees will gain insights into the architecture, implementation challenges, and measurable impacts of the system.

Running PostgreSQL in Kubernetes is becoming increasingly popular, but managing database extensions in this environment presents a challenge. Containers are designed to be immutable, making it difficult to add extensions after the database is up and running. Rebuilding containers every time you need a new extension defeats the purpose of using pre-built images with security and best practices baked in. This talk explores different approaches to managing PostgreSQL extensions in Kubernetes, including their pros and cons, and discusses potential future standards for streamlined extension management.

Stateful workloads in Kubernetes come with unique challenges that go beyond what stateless applications face. From data persistence and high availability to scalability and cost optimization, deploying stateful applications like Postgres, MongoDB, and Cassandra requires robust infrastructure and management tools. Relying on local volumes can lead to pitfalls like limited flexibility, poor resilience, and increased operational complexity.

This lightning talk explores how Software-Defined Storage (SDS) can address these challenges by providing a scalable, flexible, and cost-effective solution. Attendees will learn about the critical role SDS plays in ensuring data persistence, optimizing performance, and simplifying operations in Kubernetes environments. Through a relatable analogy and real-world insights, this talk will help Kubernetes practitioners understand the essential considerations for running stateful workloads effectively.