Hands-on Introduction to Kueue

Join Amy Chen for a hands-on introduction to Kueue, the Kubernetes-native job queueing system for batch, HPC, and AI/ML workloads.

Kubernetes knows how to schedule Pods, but running large collections of resource-intensive jobs introduces another set of challenges. Which workloads should start first? How should limited GPUs be shared between teams? What happens when demand exceeds cluster capacity?

Kueue adds workload admission, queueing, quotas, priorities, fair sharing, and preemption without replacing the Kubernetes scheduler.

In this live session, we explore:

• What Kueue is and where it fits in the Kubernetes scheduling stack
• The difference between scheduling and workload admission
• LocalQueues, ClusterQueues, ResourceFlavors, and Workloads
• Managing CPU, memory, GPU, and other resource quotas
• Sharing cluster capacity across teams and namespaces
• Priorities, fair sharing, borrowing, and preemption
• Installing and configuring Kueue
• Submitting and observing queued jobs
• Practical considerations for batch, HPC, and AI/ML platforms
• Live demonstrations and audience questions

Amy will take us beyond the architecture and show how Kueue behaves inside a real Kubernetes environment.

Whether you operate shared GPU clusters, machine-learning platforms, research infrastructure, CI systems, or large-scale batch workloads, this session will provide a practical foundation for managing queued work with Kubernetes.