Leading cloud service, semiconductor and system providers joined forces today to launch the Ultra Ethernet Consortium, a new consortium for industrywide cooperation to build a complete Ethernet-based communication stack architecture for high-performance networking.

The consortium seeks to cater to the evolving demands of artificial intelligence and high-performance computing workloads, which need top-tier functionality, performance and interoperability without compromising cost-effectiveness or usability. The Ultra Ethernet solution will use the wide applicability and flexibility of Ethernet to support diverse workloads while maintaining scalability and cost efficiency.

The founding members of the UEC include a strong lineup of well-established companies, each with the ability to provide significant contributions to high-performance solutions. The founding members are Advanced Micro Devices Inc., Arista Networks Inc., Broadcom Inc., Cisco Systems Inc., Eviden SAS, Hewlett Packard Enterprise Co., Intel Corp., Meta Platforms Inc. and Microsoft Corp.

The first chairperson of the UEC, Dr. J Metz, noted that the consortium aims not to overhaul Ethernet but to enhance its efficiency for specific workloads, examining every layer from the physical to software for optimal performance at scale.

The UEC’s mission is to foster Ethernet interoperability while minimizing communication stack changes. The consortium’s technical objectives include the development of specifications, application programming interfaces and source code.

The objectives include the development of protocols, electrical and optical signaling characteristics, link-level and end-to-end network transport protocols and telemetry and signaling mechanisms suitable for AI, machine learning and high-performance computing environments. The consortium will also work on software, storage, management and security constructs to accommodate a range of workloads and operating systems.

The UEC plans on using a systematic, modular approach to ensure compatibility and interoperability across layers, with tight integration aimed at holistic enhancement for demanding workloads. The founding companies have already contributed to four key working groups: Physical Layer, Link Layer, Transport Layer and Software Layer.

“Generative AI workloads will require us to architect our networks for supercomputing scale and performance,” Justin Hotard, executive vice president and general manager HPC & AI at HPE, said ahead of the announcement. “The importance of the Ultra Ethernet Consortium is to develop an open, scalable and cost-effective ethernet-based communication stack that can support these high-performance workloads to run efficiently.”

The UEC operates as a Joint Development Foundation project under the Linux Foundation. The consortium will start accepting new member applications in the fourth quarter of this year.

Image: Bing Image Creator