SkyRay - Extending KubeRay for Multi-Cluster Multi-Cloud Operations
CNCF [Cloud Native Computing Foundation] via YouTube
Overview
Watch a 24-minute conference talk exploring how SkyRay extends KubeRay's capabilities for multi-cluster and multi-cloud operations. Learn about Ray's unified framework for scaling AI applications and how KubeRay manages Ray clusters on Kubernetes. Discover the integration of KubeRay with a multi-cluster fleet manager, enabling seamless scheduling of Ray clusters onto cloud Kubernetes clusters based on resource requirements and policy specifications. Explore various policy configurations covering cloud providers, regions, Kubernetes versions, service quality, and GPU availability. Understand how fleet manager policy updates facilitate automatic migration of Ray clusters between Kubernetes workload clusters. Examine practical use cases demonstrating cluster selection based on resource needs, service availability, development versus production configurations, and Kubernetes version upgrades.
Syllabus
SkyRay: Seamlessly Extending KubeRay to Multi-Cluster Multi-Cloud Operation - Anne Holler, Elotl
Taught by
CNCF [Cloud Native Computing Foundation]