Coursera Spring Sale
40% Off Coursera Plus Annual!
Grab it
Learn pragmatic strategies for scaling Ray on Kubernetes through this 30-minute conference talk from Ray Summit 2025. Discover how Spencer Peterson and Raja Jadeja from Google provide a definitive playbook for using KubeRay to build stable, efficient, and self-service AI platforms on Kubernetes while eliminating operational chaos from ad-hoc ML deployments, YAML complexity, and uncontrolled resource usage. Address common ML infrastructure challenges including unschedulable pods from resource-consuming jobs, idle GPUs inflating cloud costs, developers struggling with complex Kubernetes configurations, and brittle manual deployments that don't scale. Explore how KubeRay transforms Kubernetes into an application-aware backend for distributed AI using the RayCluster custom resource for native management of Ray applications from creation and scheduling to autoscaling, fault tolerance, and cleanup. Master operational strategies for production Ray clusters including implementing true multi-tenancy with namespace isolation and cgroups for resource fairness, optimizing workload placement using Ray's placement features and Kubernetes labels for automatic job routing, reducing cloud costs through smart scaling with IPPR and Ray v2 autoscaler configuration, and managing unified multi-cluster platforms with multiple specialized Ray clusters under one Kubernetes control plane. Gain an actionable blueprint for transforming Kubernetes into a robust, cost-efficient, developer-friendly platform for distributed AI powered by KubeRay.
