AMD announced on the 12th that it has collaborated with the Energy Sciences Network (ESnet) and the US Department of Energy (DOE) to launch 'ESnet6', a high-performance network dedicated to the latest science field.

ESnet6 is equipped with an extreme-scale packet monitoring system, and AMD's FPGA-based network connection accelerator card 'Alveo™ U280' is used as the core network switching node of the system. It improves performance through high-touch packet processing that enhances network monitoring and management accuracy, and also allows for the addition of new features for continuous network innovation based on the advantages of programmable hardware.

AMD is leveraging its proprietary OpenNIC overlay to customize and embed its Alveo U280 2x100Gb/s accelerator as a network interface card (NIC) for ESnet6, providing standard network interface and host connectivity hardware to easily implement new and experimental networking features on Alveo cards.

OpenNIC, available as open source, is being utilized not only by ESnet but also by various leading academic research groups. Furthermore, AMD supports the AMD VitisNetP4 development tool, which compiles the P4 packet processing language to FPGA hardware, enabling non-hardware experts to quickly implement these technological innovations.

AMD's Alveo U280 cards, leveraging OpenNIC and VitisNetP4, are deployed across nodes on the ESnet6 network. A high-touch approach based on FPGA-accelerated processing provides an ultra-fast monitoring system for all packets in ESnet6, enabling precise analysis of network behavior and rapid detection and correction of network problems or hotspots.

Additionally, the Alveo U280 card with the OpenNIC platform provides exceptional adaptability to continuously support new and advanced features required by the end-user community throughout the lifecycle of ESnet6.

“AMD FPGA-based adaptive networking technology forms a key building block for ESnet6’s large-scale packet processing system and provides the foundation for DOE’s future vision for distributed remote science (e-Science),” said Ivo Bolsens, CTO and senior vice president of the Adaptive and Embedded Computing Group at AMD.

Meanwhile, AMD and ESnet have been working closely since 2018 to integrate powerful adaptive computing technologies to implement ESnet6's programmable smart network nodes and enable continuous network innovation.