Kunpeng Topology Affinity Plugin

Currently, most cloud service providers need to deploy a large number of containers on a server, and the number of deployed containers even exceeds the number of physical cores of the server. In this scenario, the Kunpeng Topology Affinity Plugin (Kunpeng TAP) can automatically adjust the pod deployment range based on the node resource usage to ensure NUMA affinity for container resources like CPUs and GPUs and prevent cross-NUMA memory access of service containers.

The Kunpeng TAP runs at the node layer of the Kubernetes cluster. It is deployed through pods that use DaemonSet (Daemon pods), which ensures a plugin instance on each node. This plugin interconnects with the NRI of containerd to dynamically adjust the CPU scheduling range of a pod, securing effective NUMA affinity.

The following figure shows how the NUMA affinity scheduling plugin works based on the NRI function of the containerd runtime.

In addition, to be compatible with Docker and containerd in earlier versions of Kubernetes clusters, the Kunpeng TAP can obtain container requests and return responses as a proxy, and optimize and adjust resource parameters during request delivery to implement NUMA affinity. The following figure shows how the Kunpeng TAP works.