Aviz ONES + NVIDIA Spectrum-X Ethernet: Intelligent Orchestration for AI and HPC Fabrics
October 8, 2025
ONES & SONiC
Executive Brief: The Challenge
AI and HPC workloads depend on fast, lossless, and scalable GPU networking. NVIDIA Spectrum-X Ethernet achieves this with Spectrum-4 switches, NVIDIA SuperNICs, and NVIDIA accelerated computing platforms.
Operating advanced AI infrastructure requires visibility to help enterprises navigate several challenges:
- Scaling Complexity – Expanding from hundreds to thousands of GPUs requires automation.
- Multi-Tenancy Demands – AI factories must securely isolate and manage multiple tenants and workloads.
- Operational Gaps – Traditional tools lack GPU-aware visibility across NICs, switches, and fabrics.
- Day 2 Burden – Drift detection, device RMA, and design compliance are error-prone when performed manually.
This is where Aviz ONES (Open Networking Enterprise Suite) comes in - complementing NVIDIA Spectrum-X Ethernet reference architectures with orchestration, lifecycle management, and operational intelligence to accelerate enterprise AI infrastructure deployments.
How NetOps Accelerates the Benefits of Spectrum-X Ethernet
Spectrum-X Ethernet establishes the high-performance AI fabric. ONES complements it with:
- Intelligent Orchestration – Intent-driven design and zero-touch deployment.
- GPU-Aware Multi-Tenancy – VXLAN/EVPN segmentation for secure tenant isolation.
- Full-Stack Observability – Real-time telemetry for GPUs, NICs, and switches.
- Lifecycle Automation – Backup/restore, drift detection, automated RMA.
"Enterprises want more than raw throughput—they want repeatable Day-0 design, frictionless Day-1 turn-up, and low-touch Day-2 operations," said Thomas Scheibe, CPO, Aviz Networks. "ONES complements NVIDIA Spectrum-X Ethernet with EVPN/VXLAN + per-tenant VRFs, GPU-aware provisioning, and NVIDIA Air–based intent verification, so teams can map GPU resources to tenant networks in seconds and keep clusters predictable as they grow."
Maximizing Your Spectrum-X Ethernet Investment
ONES ensures enterprises unlock the full value of Spectrum-X Ethernet by:
- Accelerating Deployment – From weeks to hours with AIR validation + orchestration.
- Secure Segmentation – Mapping of GPU resources to tenant networks in seconds.
- Ensuring Predictable AI Performance – Congestion-free, lossless fabrics.
- Simplifying Operations – GUI-based workflows reduce CLI complexity.
- Scaling Seamlessly – Add Scalable Units (SUs) without re-architecting.
- Integrating with Enterprise IT – Built-in hooks for ServiceNow, Slack, and CI/CD.
How It Works – ONES and Spectrum-X Ethernet Architecture and Orchestration
ONES aligns with the NVIDIA Spectrum-X Ethernet Reference Architecture (RA 1.3):
- Compute – HGX Hopper GPUs for AI performance.
- East-West Fabric – Spectrum-4 switches and NVIDIA SuperNICs for GPU interconnects.
- North-South Fabric – NVIDIA BlueField-3 DPU or NVIDIA ConnextX NICs for CPU, storage, and external access.
- Scalable Units – Modular 32-GPU blocks scaling to 8192 GPUs.
Where ONES complements Spectrum-X Ethernet:
- Day-0 – Intent-based design, NVIDIA Air simulation, zero-touch provisioning, and intent verification for green-field and brownfield environments
- Day-1 – Multi-tenant provisioning and GPU-aware policy orchestration, API integration with service provisioning tools
- Day 2 – Expansion, anomaly detection, RMA automation, enterprise workflow integration.
Conclusion: The Aviz Advantage in Action
NVIDIA Spectrum-X Ethernet provides the high-performance foundation. Aviz ONES completes the solution with intelligent NetOps. Purpose-built for GPU-centric fabrics, ONES delivers the orchestration, observability, and lifecycle automation needed to turn Spectrum-X Ethernet into a truly enterprise-ready multi-tenant AI fabric.
