numa-affinity

The NUMA affinity extension targets performance-sensitive workloads on multi-NUMA architectures. Through fine-grained lifecycle resource management, it significantly reduces cross-NUMA memory access and communication latency. During cluster scheduling, topology-aware policies select nodes to preserve NUMA affinity for Pods. At allocation time, optimal NUMA Distance strategies minimize cross-NUMA overhead. At runtime, Pods with frequent network interactions are consolidated onto the same NUMA node. Together with intuitive NUMA topology monitoring and container allocation visualization, the component forms a closed loop from intelligent scheduling and allocation to dynamic optimization and observability.