Aviz ONES, NVIDIA Spectrum-X Ethernet and Rafay outline a governed multi-tenant GPU platform

Aviz, NVIDIA, and Rafay describe a joint stack that pairs Aviz ONES with NVIDIA Spectrum-X Ethernet and the Rafay Platform to manage GPU infrastructure as a governed, multi-tenant AI platform, with automation spanning fabric setup through Kubernetes deployment.

Research Overview

The vendor brief frames the problem as low utilization of GPU hardware when enterprises lack an operational model for provisioning, access, and governance.

It presents the combined solution as a way to convert static GPU infrastructure into a cloud-native, multi-tenant platform that developers and data scientists can consume.

Key Findings

The brief states that the approach targets 50–70% higher GPU utilization through automated allocation and policy-driven multi-tenancy.

It also links the stack to end-to-end automation and self-service access backed by governed catalogs and chargeback.

Technical Breakdown

The stack is described as including automated fabric-related functions, positioned as lossless Ethernet fabric for AI workloads using NVIDIA Spectrum-X Ethernet.

On the software side, the brief says it supports Kubernetes cluster deployment and self-service GPU access under policy controls.

Operational Impact

The brief characterizes the operational model as covering the “Day-0 to Day-2” AI lifecycle across network, compute, and workloads.

It also describes governance mechanisms that include multi-tenant support, policy-driven GPU allocation, and chargeback to manage usage within the platform.

Overall, the article ties Aviz ONES, NVIDIA Spectrum-X Ethernet, and the Rafay Platform to an automated, governed multi-tenant operating model for GPU fabrics, from provisioning through Kubernetes deployment and controlled developer access; this “Blog Signals brief” is a fact-based summary of the vendor blog.