Loading video player...
GitOps for HPC: Bringing Cloud-Native DevOps Practices to High Performance Computing Environments - Pavan Madduri, W.W.Grainger High Performance Computing environments have traditionally relied on manual deployment processes and specialized tooling. As HPC converges with AI/ML workloads and cloud-native technologies, there's a compelling opportunity to adopt proven DevOps practices from the Kubernetes ecosystem. This session shares practical experiences applying GitOps workflows and Kubernetes orchestration to demanding computational environments. Drawing from production deployments using ArgoCD, Crossplane, and KEDA (Kubernetes Event-driven Autoscaling), I'll demonstrate how infrastructure-as-code and continuous delivery patterns improve software productivity in HPC pipelines. Key takeaways: • Implementing GitOps-driven continuous delivery for HPC workloads • Leveraging Kubernetes operators and Crossplane for infrastructure provisioning • Event-driven autoscaling patterns for burst computational jobs • Observability strategies with OpenTelemetry for distributed applications • Lessons learned from CNCF contributions applied to HPC scenarios Attendees will leave with actionable patterns for modernizing their HPC software delivery pipelines while maintaining performance and reliability requirements.