Admission and quota (Kueue ClusterQueue / LocalQueue / cohorts).
Topic hub
GPU scheduling on Kubernetes.
GPU scheduling on Kubernetes is how scarce accelerators are admitted and placed: kueue vs volcano for quota versus gang scheduling, kueue gpu ClusterQueue and LocalQueue cohorts, and gpu scheduling kubernetes priority lanes for inference and training.
01
Coverage
What this hub covers.
GPU scheduling on Kubernetes: quota and admission with Kueue, gang scheduling with Volcano or KAI, and how they fit with inference workloads.
GPU admission control with Kueue quotas, cohorts and preemption is work in progress; repository and evidence will be linked when published.
Gang scheduling and placement (Volcano, KAI).
Inference vs training queues and priority lanes.
Preemption and fair sharing.
02
Guides
Related guides.
Kueue vs Volcano vs KAI Scheduler for GPU Workloads
How Kueue, Volcano and NVIDIA KAI Scheduler differ for GPU inference and training on Kubernetes: architecture, queues, quotas and when to pick which.
03
Related cases
Evidence nearby.
LLM Infrastructure Runtime
LLM workloads move faster than traditional platform controls and can quickly become expensive, opaque, and hard to operate
LLM serving with operability signals
AI Runtime Platform
AI workloads often start as API calls but become production systems that need identity, policy, scaling, telemetry, and clear ownership
Every model call: policy + trace + budget
04
Related hubs
Nearby topics.
Available for meaningful infrastructure conversations