Shaun Bathgate
Senior Advanced Research Computing Systems Administrator, University of Victoria
Traditional high-performance computing (HPC) environments have served as the backbone for large-scale, data-intensive research for decades, using specialized schedulers and tools like Apptainer/Singularity to run containerized jobs. Yet, researchers often encounter gaps when looking to host interactive applications, automate ephemeral workloads, or facilitate rapid collaboration.
Enter Kubernetes: a cloud-native platform originally designed for web-scale deployments, now increasingly complementing—not replacing—HPC. In this session, we will demystify the misconception that “HPC already has container orchestration” by highlighting the unique strengths Kubernetes brings to research computing. We will illustrate how HPC excels at massive parallelization and batch job scheduling, while Kubernetes shines in managing microservices, on-demand interactive workloads, and event-driven pipelines.
Attendees will learn that embracing Kubernetes does not burden HPC users with complex new responsibilities—platform teams manage cluster operations behind the scenes, allowing HPC practitioners to keep focusing on building container images and running jobs. We will also address the common trope that “Kubernetes is overkill,” emphasizing how a well-managed cluster environment abstracts away the complexities of platform provisioning and extends the HPC workflow to accommodate dynamic, service-oriented computing.
By the end of this talk, attendees will understand how Kubernetes and HPC schedulers can peacefully coexist and even reinforce each other, ultimately fueling more flexible, scalable, and innovative research computing for Canadian post-secondary institutions and partners.