THE LINUX FOUNDATION PROJECTS

The New Era Needs a New Network

Our Mission

Deliver an Ethernet based open, interoperable, high performance, full-communications stack architecture to meet the growing network demands of AI & HPC at scale

Latest NewsDownload 1.1 SpecificationRead 1.0 White Paper

Steering Members

AMD
ARISTA
Broadcom
Bull
Cisco
Hewlett Packard Enterprise
Intel
Meta
Microsoft
Oracle

General Members

Alibaba Cloud
Arrcus
Astrape Networks
ByteDance
Cadence
Cornelis Networks
Dell Technologies
Everpure
Google Cloud
Huawei Technologies Co., Ltd
Keysight
Lawrence Livermore National Laboratory
Lenovo
Marvell
MIPS
New H3C Technologies Co. Ltd.
Nokia
NTT
Nvidia
Preferred Networks, Inc.
Qualcomm, Inc.
Solidigm
Synopsys
Viavi Solutions
ZTE

Contributor Members

About Us SmartDV
ALPHA NETWORKS Inc.
Amphenol Corporation
ARM
Astera Labs
Asterfusion
Centec Communications
Ciena
Clounix
Credo
d-Matrix
DRIVENETS
Edgecore Networks
Ericsson
Eridu
Extreme Networks
Fujitsu Limited
Furiosa
Grovf
Hangzhou Xinqi Electronic Co., Ltd.
Hedgehog
Intellectual Highway, Corp.
Internet Initiative Japan
IP Infusion
Kalray
Kiwimoore
Lightmatter Inc.
Linkdata
Los Alamos National Laboratory
MangoBoost
Micas Networks
Microchip
Micron Technology
Montage Technology
Napatech
Nebula Metrics
Numascale
QingHong Electronics
Qumulo
Realtek
Ruijie Networks
Sambanova Systems Inc
SAMSUNG SDS
Sandia
Scala Computing
Shenzhen Jaguar Microsystems Co. Ltd
SudoInfo Tech SDTech
Supermicro
Supranett
Teledyne
Tencent
Toyota
Ufispace
UNIVISTA
UpscaleAI
VDURA
Vultr
Xsight Labs

*Please note that not all members are displayed on this page.

Frequently Asked Questions

  • Deliver a complete architecture that optimizes Ethernet for high performance AI and HPC networking, exceeding the performance of today’s specialized technologies.  UEC specifically focuses on functionality, performance, TCO, and developer and end-user friendliness, while minimizing changes to only those required and maintaining Ethernet interoperability.
  • Additional goals: 
    • Improved bandwidth, latency, tail latency, and scale, matching tomorrow’s workloads and compute architectures.
    • Backwards compatibility to widely-deployed APIs and definition of new APIs that are better optimized to future workloads and compute architectures.

  1. How does UEC differ from what is available today in Ethernet?  AI and HPC workloads present new challenges to the network such as the need for higher scale, higher bandwidth density, multi-pathing, fast reaction to congestion, and inter-dependency on the progress of individual flows (where tail latency is the figure of merit). UEC specifications will be designed to bridge these gaps and provide the increased scale required by these workloads.  UEC is targeting a complete communications stack, addressing technical issues that traverse many protocol layers and delivering ease of configuration and management.
  2. How does UEC differ from what is available today in with other protocols?  Existing protocols may address some of the above aspects (e.g., older era congestion management), but since they were designed for general purpose networks, they lack functionality critical to AI and HPC such multi pathing and ease of configuration.  Existing protocols may also be brittle in the scenarios of interest. Together with the vast UEC members’ experience with AI and HPC workloads deployed at scale, UEC will provide a compelling, well rounded solution ushering in new hardware and software products not available with Ethernet or any other networking technology today!

  1. HPC at scale
  2. AI at scale (“GPU scale-out network”)
  3. Cloud AI and HPC

It is the intention of UEC to fully engage with the relevant SDOs (Standards Development Organizations) to make UEC technology an official element of the broadly deployed family of Ethernet based technologies and products.

UEC will engage in technology conferences, webinars, and liaisons with relevant SDOs.  UEC may also pursue compliance and interop programs.

Any organization, vendor or implementor who develops hardware or software for AI and HPC or runs these workloads and would like to contribute to or take advantage of the Ethernet ecosystem as enhanced by UEC.

Yes, UEC has steering members, general members, and contributors. More information can be found further down this page.

UEC is organized as part of the Linux Foundation JDF and enjoys the status of an International Standards Organization (ISO). UEC has members that are multinational and are centered outside of the United States. UEC will be open to global participation.

UEC technical specifications and supporting documentation are publicly available on the UEC website. The current release, UE Specification 1.1, is available for download.

Additional specifications and supporting documents are available in the Technical Publications section. Information about previous releases is available in the Specification History section.

More information is available by subscribing on the UEC website (ultraethernet.org).

Product availability and implementation timelines vary by manufacturer. For information about products incorporating UEC technologies, please consult participating vendors and their official product announcements.

Buying UEC-compliant products will be similar to the ways Ethernet-based prodcuts are procured today. Existing and future UEC members have well established routes to market to allow interested parties to buy UEC compliant equipment when available.

There are multiple advantages. One of the key advantages is that UEC compliant Ethernet based equipment will be available from an open ecosystem of manufacturers.

Subscribe to the UEC Newsletter!

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