Unlocking Bare-Metal Performance: Running Windows VMs Inside Kubernetes with KubeVirt

Introduction: The Convergence of Containers and Virtual Machines

For years, enterprise IT architectures have been divided into two distinct silos: traditional Virtual Machines (VMs) for legacy applications and containers for modern, cloud-native microservices. Managing these disparate environments often requires separate tooling, infrastructure, and engineering skillsets, leading to operational inefficiencies and increased overhead.

Enter KubeVirt, an innovative open-source technology that redefines infrastructure management by allowing you to run and manage virtual machines within a Kubernetes cluster. By deploying KubeVirt on Bare-Metal VPS, organizations can achieve near-native hardware performance, enabling resource-intensive workloads—such as Windows-based enterprise applications—to run smoothly right alongside Docker containers.

What is KubeVirt and How Does It Work?

KubeVirt is a Cloud Native Computing Foundation (CNCF) incubating project that extends Kubernetes by adding virtualization capabilities. Instead of relying on a traditional hypervisor layer separate from your container orchestrator, KubeVirt integrates directly into the Kubernetes API.

Under the hood, KubeVirt schedules a VM inside a standard Kubernetes Pod. It utilizes KVM (Kernel-based Virtual Machine) technology within the container to deliver hardware-accelerated virtualization. To the Kubernetes control plane, the VM looks and behaves like any other Pod, meaning you can leverage native Kubernetes features such as:

Scheduling and Placement: Using taints, tolerations, and node affinities.
Networking: Integrating with standard Kubernetes CNI (Container Network Interface) plugins.
Storage: Utilizing Persistent Volume Claims (PVCs) via CSI (Container Storage Interface) drivers.
Monitoring and Logging: Consolidating metrics using Prometheus and Grafana.

The Power of Bare-Metal VPS for Virtualization

While it is technically possible to run Kubernetes on nested virtualization (VMs inside VMs), doing so introduces significant performance degradation, particularly for demanding operating systems like Windows Server or Windows 10/11 desktop environments.

Deploying KubeVirt on a Bare-Metal VPS eliminates the nested virtualization bottleneck. Because the Kubernetes worker nodes have direct access to the underlying physical CPU (with Intel VT-x or AMD-V extensions), RAM, and NVMe storage, the Windows VMs running inside KubeVirt achieve near-bare-metal performance. This is crucial for applications that require low latency, intense disk I/O, or heavy computational throughput.

Why Run Windows VMs Inside Kubernetes?

Migrating to microservices is a long-term journey, and many enterprises possess critical legacy systems that cannot easily be containerized. Common examples include proprietary .NET Framework applications, Active Directory domain controllers, SQL Server databases, or desktop workloads requiring specific Windows backgrounds.

Integrating these Windows VMs into your Docker/Kubernetes ecosystem provides several distinct business advantages:

Unified Operations: Manage both your containerized frontend apps and legacy Windows backend systems using a single set of manifests (YAML) and a unified deployment pipeline (GitOps).
Resource Optimization: Maximize the utilization of your Bare-Metal hardware by packing both containers and VMs onto the same nodes based on real-time resource availability.
Simplified Networking: Containers and Windows VMs can communicate securely over the internal cluster network using standard Kubernetes Services, eliminating the need for complex external routing or VPNs.

Step-by-Step Architecture: Running Windows smoothly with KubeVirt

To successfully run a smooth, high-performing Windows environment inside KubeVirt, specific architectural configurations are highly recommended.

1. Enabling Hardware Acceleration

Ensure that your Bare-Metal worker nodes have hardware virtualization enabled in the BIOS. Within KubeVirt, you must verify that the /dev/kvm device is accessible to the pods. This ensures that the Windows OS utilizes direct hardware execution rather than slow software emulation.

2. Leveraging VirtIO Drivers

Crucial Performance Tip: Standard Windows installation media does not natively include optimized drivers for virtualized storage and networking. To prevent sluggish performance or disk recognition issues, you must attach the upstream VirtIO drivers during the installation process.

KubeVirt allows you to mount the VirtIO ISO alongside your Windows installation ISO. Utilizing VirtIO drivers for your storage controllers (viostor) and network interfaces (NetKVM) dramatically improves disk read/write speeds and network throughput, matching bare-metal expectations.

3. Optimizing Storage with Local NVMe or High-Speed CSI

Windows operating systems are notoriously storage-intensive, particularly during boot sequences and system updates. For a smooth user experience, back your KubeVirt Persistent Volumes (PVs) with local NVMe drives using a high-performance local storage provisioner, or utilize a low-latency cloud-native storage solution like Ceph or Longhorn configured with ReadWriteOnce block mode access.

A Sample KubeVirt Windows VirtualMachine Manifest

Defining a Virtual Machine in KubeVirt is entirely declarative. Below is a conceptual example of a VirtualMachine Custom Resource Definition (CRD) configured for a high-performance Windows deployment:

apiVersion: kubevirt.io/v1
kind: VirtualMachine
metadata:
  name: win2022-server-vps
spec:
  running: true
  template:
    metadata:
      labels:
        kubevirt.io/domain: win2022-server-vps
    spec:
      domain:
        cpu:
          cores: 4
          model: host-passthrough
        resources:
          requests:
            memory: 8Gi
        devices:
          disks:
          - name: datavolume
            disk: {}
          - name: virtio-drivers
            cdrom: {}
          interfaces:
          - name: default
            masquerade: {}
        machine:
          type: q35
      networks:
      - name: default
        pod: {}
      volumes:
      - name: datavolume
        persistentVolumeClaim:
          claimName: windows-os-pvc
      - name: virtio-drivers
        containerDisk:
          image: kubevirt/virtio-container-disk

In this manifest, notice the use of host-passthrough for the CPU model, which allows the Windows guest OS to fully exploit the advanced instruction sets of your Bare-Metal CPU.

Conclusion: Embracing the Future of Infrastructure Modernization

Deploying KubeVirt on Bare-Metal VPS offers an elegant, powerful solution to the hybrid infrastructure dilemma. It eliminates the arbitrary boundary between containers and virtual machines, allowing DevOps teams to consolidate control planes without sacrificing the raw, low-latency performance that heavy Windows workloads demand.

By bringing Windows VMs directly into your Kubernetes Docker cluster, you preserve legacy investments, simplify operational workflows, and take a definitive step toward a fully unified cloud-native future.