Benchmarking KubeVirt Performance: Unleashing virtbench

In the world of Kubernetes, understanding the performance of your virtual machines is essential. KubeVirt enables you to run VMs alongside containers, but without proper benchmarking, you risk poor performance and resource allocation. virtbench addresses this need by providing an open-source CLI framework that executes reproducible stress tests across KubeVirt-enabled clusters, allowing you to gain insights into your infrastructure's capabilities.

virtbench operates using a client-side orchestration model combined with an in-cluster helper pod. It begins by submitting VirtualMachine objects to the Kubernetes API server, which subsequently creates the necessary DataVolume and PVC resources via the CSI driver. The CLI then tracks the VMI status through a state machine, monitoring the transition from Pending to Running. To ensure the guest network stack is operational, an ssh-test-pod deployed inside the cluster continuously issues TCP probes to each VMI's assigned IP address. The measurement concludes only when a successful TCP handshake is achieved, confirming network accessibility.

When using virtbench, be aware that a future release will include in-VM fio tooling for I/O benchmarking from inside the guest OS, which will enhance your testing capabilities. This tool is vital for understanding how your VMs behave under load, particularly during burst capacity scenarios or live migrations. Keep an eye on updates to maximize your benchmarking efforts.