NVIDIA makes the Run:ai platform open-source

NVIDIA announced that the full Run:ai platform is available as open-source software and described the release as an expansion of its DSX AI Factory Platform.[1] The company characterizes Run:ai as providing production-grade AI workload and GPU orchestration capabilities.[1] NVIDIA says the platform has helped organizations run AI workloads on Kubernetes at scale, dynamically allocating GPUs from data processing and training through inference.[1] It says the open release could help DSX partners and AI clouds increase workload throughput, improve GPU utilization and operate shared infrastructure more efficiently.[1]

The complete Run:ai source code is available in the DSX AI Factory GitHub organization, while NVIDIA also offers the integrated open-source platform through NGC as a Community Product with a Helm installation chart.[1] NVIDIA says access through NGC requires neither a paid Run:ai subscription nor a NVIDIA support entitlement, but enterprise support is not included.[1] Downloading the NGC Community Product requires acceptance of the applicable NVIDIA software license and AI product terms.[1] For Enterprise AI Factory deployments, NVIDIA says Run:ai remains available through supported commercial offerings, including standalone, Mission Control, AI Enterprise and select cloud marketplaces.[1]

NVIDIA says the release includes KAI Scheduler, GPU Fractioning Operator, KAI Resource Management and Workload Map as independently adoptable components.[1] KAI Scheduler is the Kubernetes-native scheduling engine behind Run:ai; NVIDIA released it under the Apache 2.0 license and contributed it to the CNCF as a Sandbox project.[1] It supports fair-share and quota allocation and can run in an existing cluster without the complete platform.[1] NVIDIA says the CUDA driver enforces a hard GPU-memory limit for each workload, while GPU Fractioning Operator provides Kubernetes integration and a metrics exporter showing per-pod fractions in use; the operator is scheduler-agnostic.[1]

NVIDIA identified Crusoe, Mirantis, Nebius, OpenNebula Systems, Rafay, Saturn Cloud, Spectro Cloud and vCluster as DSX ecosystem partners integrating the platform.[1] NVIDIA says Nebius offers Run:ai as a one-click application deployment in Nebius AI Cloud.[1] Crusoe says its CMK manages the Kubernetes and GPU infrastructure beneath Run:ai's scheduling and GPU sharing, and that Crusoe and NVIDIA validated deployment for cluster registration, distributed multi-node PyTorch training, fractional GPU sharing and inference serving.[3] Crusoe welcomed the open-source release, saying it offers customers more transparency, extensibility and opportunity to participate in the community.[3]