Enhancing Performance with Cornelis’ CN500: The Future of AI Networking

Cornelis Networks has introduced the CN500 networking fabric, designed for an AI-driven world, capable of connecting up to 500,000 computers or processors with zero added latency. This architecture represents a new advancement alongside Ethernet and InfiniBand, aiming to enhance the performance of AI and high-performance computing (HPC). Cornelis claims its technology outperforms the latest InfiniBand by processing twice as many messages per second with 35% lower latency, and it communicates six times faster than Ethernet protocols. The architecture addresses challenges such as traffic congestion and latency with a dynamic adaptive routing algorithm and credit-based flow control. This ensures efficient message routing even in cases of short-term congestion and allows uninterrupted application performance even if a server fails. The CN500 product consists of network cards that integrate into servers, targeting organizations looking to upgrade for AI and HPC needs. In a competitive landscape, efficient AI adoption becomes crucial for future success.

Source link