NVIDIA Open-Sources Tools for Managed Kubernetes on GPU Servers
NVIDIA has released NVIDIA Kubernetes Engine (NVKE) and NVIDIA Kubernetes Recipes (NVKR) as open-source projects under the Apache 2.0 license. Both are in public preview and were announced as parts of the NVIDIA DSX AI Factory Platform. NVKE provides the management plane for a managed Kubernetes service, while NVKR builds the operating-system images, container images, packages and other artifacts required by clusters.
The software is designed for NVIDIA GPU bare-metal infrastructure provisioned through NVIDIA Infra Controller (NICo). NVKE uses upstream Kubernetes, kubeadm and Cluster API, and gives operators an API and the nvkectl command-line tool. Each workload cluster has its own control plane, etcd and node pools. Node admission can require a TPM-backed SPIFFE identity through SPIRE, and per-cluster etcd encryption is available.
A local evaluation requires six CPUs, 12 GiB of RAM and 100 GiB of disk, without a GPU. The projects have been conformance tested on Kubernetes 1.36. However, the installation’s NVKE API uses unauthenticated HTTP on port 30080, and documented management-plane upgrade and backup procedures are absent. NVIDIA Cloud Partners Firmus and SK Telecom are integrating the projects, while several other companies are adding components to related offerings.
Written by Sakhanda Wire AI from reporting by MarkTechPost. Facts, quotes and full context are in the original article below.