OpenNebula Integrates NVIDIA Spectrum-X for AI Cloud Deployments

- The NVIDIA Spectrum-X platform is engineered to solve networking bottlenecks in large-scale AI workloads, which are often hampered by the latency and congestion of traditional Ethernet. It combines Spectrum-4 switches, capable of 51.2 terabits per second (Tbps) of switching capacity, with BlueField-3 SuperNICs (network interface cards) that provide up to 400 Gbps connectivity to servers. - A key feature of this integration is enabling secure, high-performance multi-tenancy. This allows multiple users or teams to share a common pool of expensive GPU resources efficiently, using OpenNebula to isolate workloads and prevent one job from degrading the performance of another, an issue often called the "noisy neighbor" problem. - The platform supports direct passthrough for GPUs and SuperNICs, a technique that reduces overhead by allowing virtual machines to access the physical hardware directly. This is critical for performance-sensitive AI training and inference tasks that rely on technologies like RDMA (Remote Direct Memory Access) to bypass the host CPU's kernel. - NVIDIA claims that the Spectrum-X platform can deliver up to a 1.7x performance improvement for AI clouds compared to traditional Ethernet fabrics. This enhancement is achieved through a combination of hardware acceleration and AI-optimized congestion control. - This integration is part of a larger trend of building "AI Factories," which are centralized, shared infrastructures for developing and running numerous AI models. The collaboration specifically supports NVIDIA's latest compute platforms, including the Grace Blackwell and Grace Blackwell Ultra systems. - Before physical deployment, customers can model, test, and validate their large-scale AI infrastructure designs using NVIDIA Air, a cloud-based simulation environment where the OpenNebula control plane is operational.