The talk was presented at OSCONF 2020 Hyderabad Virtual event, where I have discussed about CNCF sandbox project KubeVirt and its adoption into Cloud-Native ecosystem

1. Turning Virtual Machines
'Cloud-Native’ using KubeVirt

2. The Story So Far ……………….
…. How we manage existing infra workloads
 CONTAINER INFRASTRUCTURE AND ORCHESTRATION
Container Application and orchestration as provided by Kubernetes and other
managed service providers such as OpenShift, Rancher are becoming the standard for
new applications.
 VIRTUALIZED WORKLOADS
Virtualized Workloads are not going anywhere fast! Business reasons (cost, time to
market) and technical reasons (older/different operating system). Can be supported
on IaaS providers Vmware, OpenStack, BareMetal…
 CONVERGING INFRASTRUCTURE
How can we bring these two worlds closer together? 😐

3. How to run Containers and VMs side by
side? … the solution

4. Introducing KubeVirt
 Turn Kubernetes into a single orchestrator for containers and virtual machines.
 Started in 2016 at Red Hat
 Open sourced in January 2017
https://github.com/kubevirt/kubevirt
 Accepted into CNCF Sandbox 2019
 Apache 2.0 License

5. KubeVirt – the ideal Converged Solution

6. Why KubeVirt ?
 Growing velocity behind Kubernetes and surrounding ecosystem for new
applications.
 Reality that users will be dealing with virtual machine workloads for many
years to come
 Focus on building transition paths for users with workloads that will either
never be containerized :
 Technical reasons (e.g. older operating system or kernel)
 Business reasons (e.g. time to market, cost of conversion)

7. KubeVirt Features
 Virtual machines run in Pods using the existing container runtimes
 Virtual machines are managed using a custom “Kubernetes like” declarative
API (CustomResourceDefinitions)
 Integration with cluster level features:
Storage, network, services, etc.
 Integration with node level features:
CPUManager, multi-network, huge pages, etc.
 Focus on ease of use and a kubernetes-native look, feel, and behavior – We
run same commands to manage pods and virtual machines together 

8. KubeVirt Components
 virt-api-server
- serves as the entry point to kubevirt for all virtualisation related flows
and updates the virtualization related custom resource definition (CRD)
- take care of validating and defining the VM configs and actions in CRDS

9.  virt-launcher
- Uses libvirtd to manage the life-cycle process of VM.
- Takes care of managing volumes and container networking
- Provides cgroups and namespaces, which will be used to host the VMI process

10.  Virt-controller
- responsible for monitoring the VMI (CRDs) and managing the life-cycle of
associated pods
- is a virtualmachine controller that take care of the VirtualMachineInstance,
managing VMGroups

11.  virt-handler
- is a DaemonSet and act as minion node
- watches for changes of VMI object, does a state reconciliation
- provide update on the VM status, restart, stop processes.
- Report domain state and spec changes to the cluster and invoke node-centric
plugins that fulfils networking and storage requirements defined in VMI specs

12. KubeVirt Anatomy

14. .... KubeVirt Use Cases
 To run Virtual Machines to support new development
- Leverage Kubernetes-based developer flows while bringing in these VM-
based dependencies
 To run Virtual Machines to support applications that can’t lift and shift
- Vendors with appliances (customer kernels, custom kmods, optimized
workflows to build appliances, …) they want to bring to the cloud-native
ecosystem
 To run Virtual Network Functions (VNFs) and other virtual appliances
- VNFs in the context of Kubernetes are of continued interest, in parallel to
Cloud-Native Network Function exploration

15. References
• https://github.com/kubevirt
• https://kubevirt.io/

16. https://www.linkedin.com/in/schakraborty007/
itsmesumanc