12:00 PM | Welcome to KubeVirt Summit 2025! | Welcome to the 5th year of our annual KubeVirt event!
This will also be a quick community update that covers recent changes in the project, our Graduation submissions, and what's next for KubeVirt. |
12:10 PM | "Faster Backups, Less Storage: Unlocking Incremental Backups in KubeVirt" by Shelly Kagan | Tired of slow, storage-hungry full backups? This session introduces a game-changing new feature in KubeVirt that provides storage-agnostic, incremental backups for your Virtual Machines. We’ll go over the highlights of how this new capability works under the hood, from the QEMU-native Changed Block Tracking (CBT) to the new API for managing checkpoints and backup jobs.
We will also include a quick demo of the feature in action, showcasing how to initiate both full and incremental backups. We'll also cover the roadmap for future enhancements, including the "pull" backup mode, offering a glimpse into the future of VM protection. Whether you're a backup vendor, a cluster admin, or a KubeVirt user, this session is for you. |
12:30 PM | "Introduction of CCA on Arm", by Howard Zhang | The primary goal of Arm CCA (Confidential Compute Architecture) is to allow existing system software, such as hypervisors, to manage VM resources while protecting data in use. CCA introduces the concept of a confidential VM, called a Realm, which prevents the hypervisor and other privileged software from observing or modifying its contents.
In this session, I will introduce CCA technology on Arm and its role in Confidential Computing. With the KubeVirt Confidential Computing WG actively enabling confidential workloads in KubeVirt, understanding CCA is key to integrating Arm into the project and supporting secure cloud-native deployments.
|
12:50 PM | "GPU‑Accelerated Workloads on KubeVirt: Scaling ML/AI in Kubernetes", by Amandeep Singh and Shivani Tiwari | KubeVirt is redefining how we run virtual machines in Kubernetes but what happens when those VMs need GPU acceleration for demanding AI/ML workloads?
In this lightning talk, I will walk through how to enable GPU-backed virtual machines in KubeVirt, and why this approach is gaining traction for secure, scalable, and isolated inference pipelines. We will explore the differences between container-based and VM-based GPU allocation, and see how KubeVirt integrates with CNCF tools like Prometheus and Kubernetes scheduler to monitor and optimize performance. If you are looking to push KubeVirt beyond typical VM use cases and into production-ready ML/AI workloads, this session will give you the technical foundation and the inspiration to get started. |
1:00 PM | "From VMs to GitOps: Managing KubeVirt Workloads Declaratively with ArgoCD", by Neel Shah | What if managing virtual machines could be as seamless or automated as deploying a containerized application? In this talk, we shall discuss how to bring the power of GitOps to KubeVirt by managing VM workloads in a declarative fashion using tools such as ArgoCD or Flux. This talk will show how virtual machines can be treated as code with changes tracked in Git. And with ArgoCD, we can now automate all possible VM life-cycle operations such as provisioning, configuration, update, scaling, and live migration. The session will showcase real-world examples of versioning the VM manifests, enforcing consistency across different environments, and safely rolling back to a known good state. Whether you are modernizing legacy infrastructure or building a hybrid platform that mixes VMs and containers, this session will show you how GitOps principles can make the life of operation teams easier, fight drift, and bring VM management into the modern DevOps era.
|
1:20 PM | "From Agent-Based to Agentless: Scalable VM Backup & Restore for Large-Scale KubeVirt Deployments", by Ganesh Chandrasekaran, Muhammad Saheer Cheruvath, Kuruva Maddileti | Managing VM backup and restore at scale is a critical challenge for production environments, especially when operating across hundreds of data centers. In this session, we share our experience designing an agentless VM backup and restore framework for one of India’s largest telecom provider’s production environment covering 100+ geographically distributed data centers across India, supporting 3,500+ physical servers and 50,000+ virtual machines We will cover: * Agent-based vs. agentless backups: trade-offs and why we transitioned to an agentless model for KubeVirt at scale. * How to leverage KubeVirt’s APIs, snapshots, and volume integrations to implement scalable and automated agentless backups. * Lessons learned from migrating from an agent-based approach to a hypervisor-driven, API-first backup strategy. * Designing high-throughput, incremental backups that minimize impact on VM workloads in distributed clusters. * Strategies for application-aware backups using declarative automation and orchestration without installing per-VM agents. * Real-world operational lessons, performance tuning, and failure recovery strategies in nationwide telco-scale deployments. Attendees will gain practical knowledge of how to architect reliable, low-touch VM backup and restore processes in KubeVirt for production-scale deployments, reducing operational overhead while maintaining application-level data protection.
|
1:40 PM | "Crossing the border - VM goes on a BGP adventure", by Or Mergi | Good old BGP protocol, well know and widely used, allows networks to talk with each other.
OVN-Kubernetes - CNCF project - is a robust network solution for Kubernetes clusters build on top of OVN and OVS. Its latest release introduces new exciting features such as user-defined networks and BGP support, enabling new use-cases for VMs.
In this session, we will go over the new developments and how they can be utilized for building your cloud using Kubevirt:
- Network isolation for multi-tenancy
- Exposing VMs network outside the cluster (BGP).
- Stretching a UDN across multiple clusters using vrf-lite.
If you like to get updated about latest features and future development around network-isolation, multi-tenancy, and BGP support, this session is for you.
|
2:00 PM | "Taming the VM Sprawl: KCP Meets KubeVirt", by Sören Henning | This session shows how you can manage KubeVirt virtual machines across multiple Kubernetes clusters from a single control point using the cncf project kcp which provides multi-tenancy out of the box.
We’ll introduce kcp workspaces, api exports and bindings for tenancy, explain how syncing between workspaces and real kubernetes clusters works and demonstrate how this setup simplifies multi-cluster VM management and tenant isolation.
The talk includes a live demo using a kcp control plane and kubevirt clusters to manage a fleet of virtual machines, making it easy to try it out on your own.
|
2:20 PM | "Got Hyper-V? Let's Move it to KubeVirt", by Sylvain Chen & Rares Donca | How do you move established VM workloads from Microsoft Hyper-V into a modern KubeVirt environment? What tools are available, and do they actually work? How do you handle the process for dozens or even hundreds of VMs without losing your mind? This session answers those questions. Drawing from our hands-on experience migrating a variety of customer workloads (Linux, Windows, and appliances), we present a comprehensive guide to a successful cold migration. We'll explore the technical hurdles you're likely to face, assess the state of open-source migration tools, and, most importantly, show you how to use Ansible to automate the entire process for repeatable, scalable results. Don't just modernize around your VMs—bring them with you into the cloud-native world.
|
2:40 PM | "Multi-Arch VM deployment options with KubeVirt + CDI", by C. A. Fillekes & Thomas-David Griedel | Now that we have KubeVirt + CDI running and released on x86, ARM and s390x, there are a few multi-arch deployment options we can execute out-of-the-box, and some (e.g. converting VMware VMs to run on the mainframe) requires some additional tools and development. Likewise, there are some options which require no code changes at all, but which provide tremendous benefit to the user. We explore the various options of multi-arch KubeVirt, and consider the options that require the least work/risk and provide are the most benefit -- demonstrations included!
|
3:00 PM | "Decentralized live migration overview", by Alexander Wels | In KubeVirt 1.6 we introduced Decentralized live migration. This is going to be about what is this, how is it different from normal live migration. Why do we care, and what can it do for you.
|
3:20 PM | "Leveling Up Kubevirt-Manager: New Features, Better UX, and Real-World Usability", by Marcelo Parisi | Kubevirt-Manager continues to grow as a lightweight and open-source-friendly UI for managing KubeVirt environments, with a focus on usability, real-world scenarios, and community-driven improvements. Since our last update, the project has made major strides to support more flexible VM configurations, improved performance, and better workflows for developers and operators alike.
Highlights include support for BIOS/UEFI and Secure Boot, dynamic forms that adapt to user input, and dynamic tables for smoother interaction with larger environments. We've improved the disk and networking experience with support for volume modes, disk types, multiple LoadBalancer ports, and boot order settings. There’s also now full serial console integration via xterm.js, and the UI is fully functional in air-gapped environments. The project has also been updated to kubevirt.io/v1, keeping it aligned with the latest upstream API.
|
3:40 PM | "Achieving Microsecond-Precision Clock Sync for KubeVirt VMs Using Hardware Timestamping", by Satish Kumar Nagireddy | Time synchronization in virtualized environments presents unique challenges when migrating legacy applications requiring precise temporal coordination to KubeVirt clusters. This presentation explores implementing hardware-accelerated time synchronization for virtual machines running latency-sensitive workloads on KubeVirt, achieving sub-microsecond precision through SR-IOV passthrough and PTP hardware timestamping capabilities.
Traditional virtualization platforms suffer from millisecond-level clock drift that impacts distributed applications, real-time systems, and financial trading platforms. Our implementation leverages KubeVirt's device plugin framework to expose hardware timestamping NICs directly to VMs, enabling precision time protocol synchronization while maintaining cloud-native operational benefits. The solution incorporates custom device plugins for PTP-capable hardware, automated clock source configuration through admission webhooks, and monitoring integration via Prometheus metrics. Production deployments across multi-node clusters demonstrate remarkable improvements with VM-to-VM clock synchronization achieving sub-microsecond accuracy compared to default virtio-based alternatives showing 10-50ms drift. Critical workloads including distributed databases and real-time analytics platforms show 90% reduction in timestamp-related conflicts. The presentation addresses implementation challenges including hardware compatibility across heterogeneous clusters, live migration considerations for time-sensitive VMs, and integration with existing Kubernetes scheduling constraints. Our findings establish clear patterns for migrating legacy applications requiring precise time synchronization to KubeVirt, providing practical blueprints for operators managing latency-sensitive virtualized workloads at scale.
|