The Ultra Ethernet Consortium, announced on July 19, brings together cloud service providers, semiconductor manufacturers and system providers for an industry-wide collaboration to build a fully Ethernet-based communication stack architecture for high-performance networks. Their reasoning: Artificial Intelligence (AI) and High Performance Computing (HPC) workloads are evolving rapidly and have the highest demands on functionality, performance, interoperability and total cost of ownership. At the same time, user-friendliness for developers and end users must not suffer. The Ultra Ethernet Solution Stack leverages the ubiquity and flexibility of Ethernet to handle a variety of workloads while being scalable and cost-effective.

The founding members are AMD, Arista, Broadcom, Cisco, Eviden (a division of Atos), HPE,Intel, Meta and Microsoft. Together, these companies have decades of experience in networking, AI, cloud and high performance computing.

Specifications, APIs and source code

"This is not about redesigning Ethernet," says Dr. J Metz, Chair of the Ultra Ethernet Consortium. "It's about optimizing Ethernet to improve efficiency for workloads with specific performance requirements. We are looking at every layer, from the physical to the software layer, to find the best way to improve efficiency and performance."

The consortium's mission is to minimize changes to the communications stack while maintaining and promoting Ethernet interoperability.

The technical goals of the consortium are to develop specifications, APIs and source code that define the following areas:

• Protocols, electrical and optical signaling characteristics, application program interfaces (APIs) and/or data structures for Ethernet communications
• Link-level protocols and end-to-end network transport protocols to extend or replace existing link and transport protocols
• Overload, telemetry and signaling mechanisms at link level and end-to-end. All of these techniques are suitable for artificial intelligence, machine learning and HPC environments
• Software, storage, management and security concepts to facilitate a variety of workloads and operating environments

The UEC takes a systematic approach with modular, compatible, interoperable layers with tight integration to enable holistic improvement for demanding workloads. The founding companies support the consortium with contributions in four working groups: Physical Layer, Link Layer, Transport Layer and Software Layer.

The UEC is a project of the Joint Development Foundation, which is supported by the Linux Foundation. The UEC will be accepting applications for new members from Q4 2023. Further information can be found at ultraethernet.org.