multi-cluster

Armada treats multiple Kubernetes clusters as a single resource pool for GPU-intensive AI workloads, with global queue management, gang scheduling, and production-scale throughput.