AI Fabric Validation at Scale - VIAVI at MWC Barcelona 2026

Why It Matters

Validating AI fabric performance pre‑deployment reduces costly GPU inefficiencies and energy waste, accelerating reliable AI service rollout for enterprises.

Key Takeaways

• VIAVI’s AI fabric test appliance emulates GPUs for traffic validation.
• Real‑time RoCE v2 and CCL traffic generated to mimic production loads.
• Nokia switch introduces congestion, triggering ECN notifications and throttling.
• Simulated GPUs cut power use by roughly 70% versus real farms.
• Web‑based wizard UI enables quick, installation‑free configuration of complex tests.

Summary

The video showcases VIAVI Solutions’ AI data‑fabric validation platform unveiled at MWC Barcelona 2026. Using the B3 test‑center appliance, the system emulates GPU workloads and generates realistic RoCE v2 and CCL traffic, routed through a Nokia switch and connected via Amphenol 400‑Gbps DAC cables.

Key capabilities include measuring job‑completion time to gauge GPU efficiency, deliberately inducing congestion on the Nokia switch, and observing ECN notifications that trigger bandwidth throttling. The demo contrasts an ideal, congestion‑free run with a throttled scenario, highlighting how congestion inflates job times and reduces GPU utilization, a costly outcome for AI data centers.

The presenter emphasizes that simulated GPUs cut power consumption by roughly 70% compared with building a physical GPU farm, and that the solution also supports conventional routing, switching, and Ethernet testing, delivering a broader ROI. Configuration is performed through a fully web‑based, wizard‑driven UI, eliminating local installations and simplifying complex test setups.

For enterprises deploying AI workloads, VIAVI’s platform offers a repeatable, low‑cost method to validate AI fabric performance before production, mitigating risk, conserving energy, and ensuring optimal GPU usage in high‑throughput environments.

Original Description

In collaboration with Nokia and Amphenol Communications Solutions, VIAVI demonstrated AI Fabric Validation at Scale at MWC Barcelona 2026.

This demo highlights how AI data center fabrics can be validated before deployment using realistic AI traffic models and measurable performance outcomes. The demonstration features the VIAVI TestCenter B3 800G Appliance emulating GPU workloads, with traffic traversing a Nokia 7220 IXR H4 switch interconnected using Amphenol 400G QDD DAC cables, illustrating an interoperable, production‑ready AI networking environment.

By showing how congestion, tuning, and fabric behavior directly impact Job Completion Time (JCT) and GPU efficiency, the demo underscores the importance of early validation to reduce risk, improve efficiency, and enable repeatable, scalable AI deployments.