Managing clusters

Clusters are responsible for AI model inference, compute resource management, and other functions. They are divided into the following two types:

Static node cluster: Suitable for grouping servers or virtual machines as static nodes into a cluster. The control plane accesses static nodes via IP + SSH information.
Kubernetes cluster: Suitable for creating an AI cluster on an existing Kubernetes cluster. Neutree only deploys the system services required for models and inference, while compute resources are managed by native Kubernetes. The control plane accesses the Kubernetes cluster via Kubeconfig.

After a cluster is created, you can view CPU, memory, and accelerator resource usage on the cluster details page. For clusters that can report GPU card-level information, the details page also shows the GPU card model, UUID, health status, VRAM usage, and compute usage for each node.