The below article syndicated from FABRIC core team member ESNet outlines multiple ways the project accelerates science, including a use case featuring FABRIC.
Large-scale scientific collaboration depends on the seamless flow of information from its source to where it can be analyzed and understood—like a vast circulatory system transporting vital knowledge. ESnet keeps this system running and enables breakthroughs that shape our world.
Funded by the U.S. Department of Energy (DOE) Office of Science and managed by Lawrence Berkeley National Laboratory, the Energy Sciences Network (ESnet) was created in 1986 to provide dedicated high-performance network connectivity for DOE researchers.
These days, ESnet connects tens of thousands of DOE-funded researchers across the U.S. and Europe, including at all 17 national laboratories and 28 “user facilities” — DOE’s premier scientific instruments and supercomputing centers. ESnet’s network, with backbone speeds of 400 Gbps or more, enables collaborators to move enormous data sets — 1.7 exabytes’ worth total in 2024 — vast distances at lightning speed, accelerating time to discovery.
ESnet offers not only fast data, but also highly specialized services and innovative applied research, including experimental testbeds, advanced wireless prototypes for field research, AI for operations, and a three-node quantum network testbed. Thanks to its close codesign partnerships with its science users, ESnet has supported Nobel Prize-winning breakthroughs from the discovery of the Higgs boson to the co-discovery of CRISPR-Cas9, a groundbreaking genetic engineering technology.
“Great networks by their nature fade into the background, allowing science to take center stage. But at ESnet we are not just a network, we are also a partner to all the scientific enterprises across the Department of Energy — and we actively collaborate with them to accelerate their scientific research,” says ESnet Executive Director Inder Monga. “ESnet is the fabric through which they access HPC, including our new real-time raw-data streaming capabilities. We’re building infrastructure like in-network caching and gathering large data sets to support the increasing demand for AI — and we’re experimenting with AI to improve our own operations. We’re also providing the practical networking expertise to QUANT-NET’s effort to move quantum networking from the physics lab into production.”
Here are several achievements that highlight how ESnet is accelerating the nation’s science.
Increasing Data Capacity for all Major DOE Facilities
Since launching the latest iteration of its high-performance network, ESnet6, in 2022 — ahead of schedule and under budget — ESnet has upgraded the connections of its major facilities to 400 Gbps or faster: Fermi National Accelerator Laboratory now has a 2.4 Tbps connection to ESnet for its high-energy particle physics data.
Serving as a “Super-Integrator” of Labs and Facilities
ESnet has been the key link in Berkeley Lab’s several “Superfacility” projects, connecting researchers around the world to the National Energy Research Scientific Computing Center (NERSC), the primary scientific computing facility for the DOE Office of Science’s six programs, with close to 10,000 users. ESnet’s network and NERSC’s Perlmutter supercomputer together enable real-time data processing that has accelerated the process of discovery for research in nuclear fusion, how nuclei decay, the COVID virus structure. ESnet is now the essential connective tissue for the DOE’s Integrated Research Infrastructure program, inspired in part by the success of the superfacility collaborations. The three IRI pathfinder projects, focused on Light Sources, Plasma Physics and Fusion Energy, and Climate and Earth Science, all depend on ESnet to integrate the research instruments seamlessly with high-performance computing (HPC) facilities.
Enabling Global Scientific Collaborations
Leveraging its R&D 100 award-winning bandwidth reservation system, OSCARS, ESnet is a longtime strategic partner for the international research community of CERN’s Large Hadron Collider (LHC), the world’s most powerful particle accelerator, and an enabler of LHC-related computing. ESnet also plays a critical role in newer international collaborations. For example, ESnet’s network helps transport the data captured by the LSST Camera of the Simonyi Survey Telescope at the Vera C. Rubin Observatory, funded by the U.S. National Science Foundation and the U.S. Department of Energy’s Office of Science, for the Legacy Survey of Space and Time (LSST). ESnet collaborates with international partners to carry these incredibly data-rich images from Chile to the SLAC National Accelerator Laboratory in California, where astronomical object alerts are produced within 60 seconds of image acquisition.
Achieved Real-Time Streaming of Raw Physics Data
A collaboration between ESnet and Thomas Jefferson National Accelerator Facility exemplifies ESnet’s codesign approach. The EJFAT prototype has been designed to allow different types of facilities, such as X-ray light sources and particle accelerators, to stream raw data in real time to geographically distributed supercomputers for processing and analysis. In 2024, EJFAT successfully streamed physics data from Jefferson Lab at 100 Gbps into NERSC; since then, it has connected to the Oak Ridge Leadership Computing Facility and the FABRIC testbed. Real-time streaming capability means that researchers can more quickly and easily calibrate and steer time-sensitive experiments while they’re happening — saving money and time.
Leveraging AI for Infrastructure, and Infrastructure for AI
In the 2023 AI for Science, Energy, and Security Report published by the DOE Office of Science and the National Nuclear Security Administration, ESnet was mentioned two dozen times, primarily for its capacity to move large data volumes around for AI learning and to integrate the other DOE facilities. In addition to looking at how ESnet can best support the infrastructure needs of AI, ESnet is experimenting with AI to analyze, troubleshoot, and predict the network traffic flowing across this vital data circulatory system. To do so requires more than software. ESnet has installed programmable, off-the-shelf devices at 86 U.S. and European locations that collect data via precision telemetry. By applying AI queries and insights, ESnet is improving its understanding of this critical scientific network and innovating to support future data-intensive AI collaborations. It is also working with AI academic researchers to use this data to develop foundational models for networking.