Kubernetes in Action: Exploring the Performance of Kubernetes Distributions in the Cloud
Hossein Aqasizade, Ehsan Ataie, Mostafa Bastam
TL;DR
This paper evaluates how four Kubernetes distributions (Kubeadm, K3s, MicroK8s, K0s) perform when OpenFaaS is deployed as a FaaS workload on CloudLab clusters. It systematically varies underlying virtualization (PV vs HVM) via Xen and container runtimes (Docker vs Containerd) to identify the best configuration. Through IOzone, Sysbench, and OpenFaaS breakpoint tests, the study reports metrics such as throughput, latency, and scaling behavior. Key findings show that Xen-PV outperforms HVM, Docker outperforms Containerd under heavy load, and K3s offers the best balance of throughput and CPU usage, with K0s failing under stress. The results inform deployment decisions for cloud-based Kubernetes + serverless platforms and highlight directions for future optimization, including ML-guided resource management and anomaly detection.
Abstract
Kubernetes has emerged as a leading open-source platform for container orchestration, allowing organizations to efficiently manage and deploy containerized applications at scale. This paper investigates the performance of four Kubernetes distributions, namely Kubeadm, K3s, MicroK8s, and K0s when running OpenFaaS as a containerized service on a cluster of computing nodes on CloudLab. For this purpose, experiments are conducted to examine the performance of two virtualization modes, namely HVM and PV, supported by Xen as the underlying hypervisor. Moreover, two container runtimes that are integrated with Kubernetes, namely Docker, and Containerd, are examined to assess their performance on both disk-intensive and CPU-intensive workloads. After determining the appropriate underlying Xen mode and container runtime, the Kubernetes distributions are set up and their performance is measured using various metrics, such as request rate, CPU utilization, and scaling behavior.