Enterprise Reference Architecture for Object Storage with VAST Data CNodes and DBoxes
Your GPUs Are Only as Fast as Your Data Platform
Modern AI factories—driving agentic AI and complex, structured, and unstructured data pipelines—demand a high-performance data platform that exceeds traditional storage limitations. To achieve efficiency and productivity at scale, your infrastructure must integrate multi-protocol, high-concurrency access that seamlessly bridges block, NFS and S3 via the VAST AI Operating System.
The NVIDIA Enterprise Reference Architecture (ERA) for Object Storage, featuring VAST Data CNodes and DBoxes, provides a validated design to eliminate bottlenecks and reduce deployment risk. NVIDIA-Certified at the Foundation level for workloads scaling to 128 GPUs, this architecture delivers the speed, scale, and operational efficiency required to power everything from classical analytics to generative AI.
What's Inside
The NVIDIA Enterprise Reference Architecture for Object Storage with VAST Data CNodes and DBoxes is a design guide from VAST Data and NVIDIA, validated under NVIDIA-Certified Storage at the Foundation level for AI workloads scaling to 128 GPUs. It's built for storage architects and infrastructure engineers who need a proven, deployable design that addresses the rapidly growing multi-protocol storage needs of accelerated analytics, classical AI, and generative AI workloads.
Get the Full Reference Architecture
See complete compute, networking, and storage scaling guidance as you architect your next-generation AI infrastructure.