If you run training or batch AI jobs across shared clusters, this names the failure modes that come with static resource configs and hand picked clusters, and shows the Ray on Kubernetes job layer Uber built instead.

• articleUbers Journey To Ray On Kubernetes Resource Management
• articleEfficient Software Factory

• articleUbers Journey To Ray On Kubernetes Resource ManagementUber Engineering 1
• articleSkypilotShopify Engineering
• articleHow to Run Isolated Tenant Kubernetes Clusters on Shared GPU InfrastructureTanya Lenz

Checking sign-in…

Loading comments…