What's Happening?
DriveNets, a prominent player in high-scale networking solutions, has announced the first commercial deployment of an AI supercluster utilizing long-distance scale-across AI networking. This development is part of Project Redwood, initiated by WhiteFiber,
a leading AI infrastructure solutions provider. The DriveNets AI Fabric connects two WhiteFiber data centers, located over 50 miles apart, into a single logical GPU supercluster. This setup has been validated to deliver 111.2 Tbps of bandwidth with a guaranteed latency of 0.9ms. The scale-across architecture allows for the extension of AI clusters beyond the power and space limitations of a single site, enabling larger and more resilient clusters. This architecture addresses the challenges of AI training, which involves large, synchronized data flows that conventional data center networks struggle to handle.
Why It's Important?
The deployment of this AI supercluster marks a significant advancement in AI infrastructure, addressing the power and space constraints that have traditionally limited AI buildouts. By enabling the extension of AI clusters across multiple sites, the scale-across architecture allows for greater flexibility in location selection, potentially reducing costs and increasing efficiency. This development could lead to more robust AI systems capable of handling larger datasets and more complex computations, benefiting industries reliant on AI technologies. The ability to manage congestion and latency effectively across long distances is crucial for maintaining performance, making this a pivotal step in the evolution of AI infrastructure.
What's Next?
WhiteFiber plans to expand the capabilities of this AI supercluster by adding additional scale-across ports, aiming to achieve 136 Tbps of bandwidth by the third quarter of 2026. This expansion will further enhance the performance and capacity of the AI infrastructure, potentially attracting more businesses and industries to adopt similar technologies. The success of this deployment could encourage other companies to explore scale-across architectures, leading to broader adoption and innovation in AI networking solutions.
