The Linux Foundation Projects
Skip to main content

Working Groups

Are you ready to embark on a journey into the future of high-performance networking? The Ultra Ethernet Consortium (UEC) is your gateway to a world of cutting-edge technology, where we’re redefining Ethernet for the demands of Artificial Intelligence (AI), and High-Performance Computing (HPC).

At UEC, our mission is clear: to develop Ethernet specifications and source code that empower AI, and HPC environments with next-level performance, scalability, and interoperability. We’re paving the way for seamless data exchange and computation in the digital realm, and we want you to be a part of this exciting transformation.

Our Pillars: Working Groups

UEC is structured around dynamic Working Groups, each with a unique focus on different aspects of Ethernet technology. Together, these groups are reshaping the Ethernet landscape to meet the evolving needs of AI, and HPC environments. Let’s dive into each one:

Physical Layer

In the world of Ethernet, the Physical Layer is the foundation upon which everything else rests. Our Physical Layer Working Group is dedicated to enhancing the performance, reducing latency, and improving the management of Ethernet’s physical infrastructure. This includes the development of specifications for the Ethernet physical layer, electrical and optical signaling characteristics, application program interfaces, and data structures. Our goal is to make the foundation stronger, ensuring that Ethernet can meet the rigorous demands of AI, and HPC.

Link Layer

The Link Layer is where data packets are organized for efficient communication. Our Link Layer Working Group is all about enhancing the performance, reducing latency, and improving the management of Ethernet’s link layer. We develop specifications that optimize Ethernet’s efficiency, security, and scalability. We’re extending and replacing existing link and transport protocols to ensure data flows smoothly and securely in AI, and HPC environments.

Transport Layer

The Transport Layer is crucial for end-to-end data delivery. Our Transport Layer Working Group focuses on developing specifications for an AI/HPC transport that delivers enhanced throughput, reduced latency, greater scalability, and improved management in Ethernet networks. We’re ensuring that Ethernet can handle the high-performance demands of AI, and HPC applications without missing a beat.

Software Layer

In the age of software-defined everything, our Software Layer Working Group takes center stage. We’re developing specifications, software APIs, and open-source code to support a wide range of AI/HPC use cases and applications. This includes optimizing remote memory access, enabling In Network Computations (INC), and addressing security, management, and storage concerns. We’re making Ethernet more flexible, programmable, and adaptable for AI, and HPC environments.

Storage Working Group

The Storage Working Group adds storage services to Ultra Ethernet (UEC) based AI and HPC workloads, collaborating with other UEC workgroups for seamless integration. Our focus spans compatibility with industry-standard physical connections, link-level capabilities negotiation, and packet encoding mechanisms as well as addresses emerging storage services. We integrate transport services, enhance security, and ensure RDMA API compatibility through collaboration with the software workgroup. Finally, we address the challenging task of integrating industry-established practices of storage management with the UEC, bolstering the foundation for high-performance storage traffic.

Compliance Working Group

The Compliance Working Group is focused on UEC specifications for AI and HPC workloads and is all about making sure services and devices meet the UEC defined technology. We define specs and create tests to assess UEC implementations, ensuring rigorous compliance with ratified UEC standards. We also define interoperability goals among UEC compliant network devices (e.g., integrated NICs, PCIe NICs and switches) according to UEC defined AI and HPC profiles.

Management Working Group

The Management Working Group enhances UEC fabric manageability by collaborating with other workgroups to integrate evolving management specifications. We define management elements, RPCs, and models for switch and UEC compliant Transport Fabric End Point (FEP) management. Our efforts include topology discovery, capability discovery, monitoring, and interoperability queries to ensure seamless interaction among UEC fabric components.

Performance and Debug Working Group

The Performance and Debug Working Group defines performance benchmarks and debug capabilities and tools for AI and HPC workloads. We align with evolving UEC specifications and define performance metrics, benchmarks, and debug features. Our aim is to create performance metrics that communicate to customers UEC’s capabilities, performance and unique merits for their application. It will also assist developers, DevOps and network operation teams by enhancing visibility and debuggability in UEC-compliant implementations. We provide test cases and software tools for validation, creating a robust environment for AI and HPC applications

Join the Ethernet Evolution

At UEC, we invite you to be a part of the Ethernet evolution. Whether you’re passionate about hardware, protocols, transport, or software, there’s a place for you in our community. Together, we’re shaping the future of Ethernet to empower AI, and HPC, and to ensure that products across the industry can seamlessly interoperate. 

Join us today, and let’s redefine the boundaries of what Ethernet can achieve in the world of AI,  and HPC. Together, we’re connecting the future!

Become a Member

Subscribe to the UEC Newsletter!

Get the latest announcements, event info, and the community news in your inbox.