- InfiniBand's long dominance faces real pressure from the open Ethernet movement
- Meta and Nvidia are betting on openness to scale AI networks
- The ESUN project unites industry competitors through shared networking ambitions.
Open Computing Project (OCP) announced a new initiative known as Ethernet for Scalable Networking (ESUN), aimed at developing open standards for high-performance connections in artificial intelligence clusters.
This collaboration brings together companies such as Meta, Nvidia, AMDCisco and OpenAI to explore how Ethernet can compete with existing interconnects such as InfiniBand in large-scale networks. data centers.
Other companies are also joining the collaboration: Arista, HANDBroadcom, HPE Networking, Marvell, Microsoftand Oracle.
Open network for AI clusters
InfiniBand has long dominated the high-speed AI networking market, accounting for approximately 80% of interconnect infrastructure. GPUs and accelerators.
However, the ESUN team believes that Ethernet's maturity, cost-effectiveness and interoperability make it a strong candidate for scaling AI clusters.
Unlike proprietary systems, widespread adoption of Ethernet among engineers can help reduce the complexity of managing massive AI workloads.
Proponents argue that using Ethernet as an open standard will allow operators to scale infrastructure while reducing costs.
new OCP Artificial Intelligence Tools The initiative builds on earlier work in the SUE-Transport (SUE-T) program, which explored Ethernet transport for multiprocessor systems.
ESUN members will meet regularly to define standards for switch behavior, including protocol headers, error handling, and lossless data transfer.
The group will also examine how network design affects load balancing and memory ordering in GPU-based systems.
It plans to coordinate with the Ultra Ethernet Consortium and the IEEE 802.3 standards body to ensure consistency across the broader Ethernet ecosystem.
Several firms have already developed Ethernet-based products designed to scale AI: Broadcom's Tomahawk Ultra switch supports up to 77 billion packets per second, for example, and Nvidia's Spectrum-X platform also combines Ethernet with hardware acceleration for AI clusters.
However, Meta, which co-founded OCP in 2011, sees ESUN as a natural extension of its commitment to open hardware in the data center.
However, observers note that replacing existing InfiniBand networks will require Ethernet to prove itself in the most demanding AI workloads, where latency and reliability are critical.
The success of ESUN will depend on a balance of openness and productivity. Proponents envision a future in which artificial intelligence systems run on interoperable hardware using standardized Ethernet technologies.
However, given the scale and sensitivity of AI infrastructure, it remains unclear whether there will be a decisive industry shift away from proprietary interconnects.
For now, ESUN is an ambitious effort and it remains to be seen whether it can match the performance of InfiniBand.
Follow TechRadar on Google News. And add us as your preferred source to get our expert news, reviews and opinions in your feeds. Be sure to click the “Subscribe” button!
And of course you can also Follow TechRadar on TikTok for news, reviews, unboxing videos and get regular updates from us on whatsapp too much.
