Manage LLM evaluation workloads at scale with EvalHub and Kueue

Part 8 of an EvalHub series explaining how Kueue manages large-scale LLM evaluation workloads on Kubernetes through queue-based scheduling, addressing resource contention, uncontrolled consumption, and instability. It describes fair resource sharing via quotas, priority-based preemption, and automatic queueing, defines three personas (cluster administrator, namespace owner, ML engineer), and covers operator installation, ClusterQueue and LocalQueue configuration, and job submission via API.