en.Wedoany.com Reported - The Nutanix Unified Storage (NUS) solution has recently obtained NVIDIA enterprise certification, helping enterprises and cloud providers deploy storage infrastructure capable of supporting large-scale production-grade AI workloads. Thomas Cornely, Executive Vice President of Product Management at Nutanix, stated that this certification validates NUS's ability to deliver the full-stack interoperability, linear scalability, and reliable data transfer speeds required by modern AI workloads. Jason Hardy, Vice President of Storage Technology at NVIDIA, noted that as enterprises expand AI factory deployments, storage becomes the foundation for unlocking full-stack performance, efficiency, and accuracy. With this certification, NUS's full-stack interoperability within NVIDIA's AI infrastructure is validated, helping to reduce I/O bottlenecks and integration risks. Nutanix also announced plans to support NVIDIA Vera BlueField-4 STX, reinforcing its commitment to enhancing data access speed, storage efficiency, and simplified management in large-scale AI operations.

When building AI factories, enterprises and cloud providers require infrastructure that can sustain data flow, maximize GPU utilization, and reduce deployment risks. Fragmented infrastructure, data silos, and inconsistent performance are key factors slowing down deployment, limiting GPU efficiency, and hindering reliable AI scaling. With the NVIDIA certification, NUS provides a validated configuration option, supporting enterprise-grade AI infrastructure deployments.

Based on a 10-node all-NVMe cluster, NUS leverages enhanced parallel NFS (pNFS) and GPUDirect Storage via NFS over RDMA to establish a low-latency, high-throughput, and resilient data path between GPUs and storage. This solution utilizes NVIDIA Spectrum-X Ethernet, including Spectrum-4 switches and BlueField-3 DPUs, achieving linear scaling from 10 GB/s read and 5 GB/s write with 32 GPUs to 160 GB/s read and 80 GB/s write with 1,024 GPUs.

This resilient zero-downtime architecture supports training, fine-tuning, inference, and RAG pipelines, covering compute platforms including x86-based systems (NVIDIA RTX 6000 PRO Blackwell, NVIDIA H200 NVL), NVIDIA HGX servers equipped with B200, H200, or H100 GPUs, and the NVIDIA GH200 Grace Hopper superchip.

The NVIDIA-certified Nutanix Unified Storage reference architecture is now available. Planned support for NVIDIA BlueField-4 STX is expected to be available in the second half of 2026.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com