Home // CLOUD COMPUTING 2025, The Sixteenth International Conference on Cloud Computing, GRIDs, and Virtualization // View article

Running Kubernetes Workloads on Rootless HPC Systems using Slurm

Authors:
Jonathan Decker
Sören Metje
Julian Kunkel

Keywords: Kubernetes, HPC, Container, Slurm, Cloud

Abstract:
Kubernetes has become a widespread orchestrator for cloud workloads but with increasing demand for compute the need arises to also access HPC environments that are operated via batch schedulers such as Slurm. A number of solutions for combining Slurm and Kubernetes are available, which can be categorized further based on the interaction between Slurm and Kubernetes that they provide. In this paper, we consider the use case of utilizing an existing Slurm cluster to run Kubernetes workloads. For this, we introduce a new solution called Kind Slurm Integration (KSI) based on Kind and rootless Podman and compare it based on performance, usability and maintainability to the existing solutions Bridge Operator and High-Performance Kubernetes (HPK). We found that Bridge Operator provides native performance as it effectively submits Slurm jobs through a Kubernetes interface and that HPK provides good performance by creating almost feature complete Kubernetes clusters on top of Apptainer. KSI on the other hand is able to provide fully functional Kubernetes clusters inside Slurm jobs but lacks behind in network performance. Overall, we conclude that more work is needed to run Kubernetes workloads under Slurm without missing out on features or performance.

Pages: 100 to 107

Copyright: Copyright (c) IARIA, 2025

Publication date: April 6, 2025

Published in: conference

ISSN: 2308-4294

ISBN: 978-1-68558-258-6

Location: Valencia, Spain

Dates: from April 6, 2025 to April 10, 2025