Organizers:Ilya Baldin (UNC RENCI)
James Griffioen (University of Kentucky)
Inder Monga (ESnet)
Anita Nikolich (University of Illinois)
Kuang-Ching Wang (Clemson University)
Speakers:James Griffioen (University of Kentucky) -- FABRIC Overview
Ben Kirtman (University of Miami) -- Scientific Computing
Jason Mancuso (Cape Privacy) -- Distributed Machine Learning
Christos Papadopoulos (Colorado State) -- IoT and Scientific Devices
Ilkay Altintas (UC San Diego) -- Data Analysis and Workflows
Panel Abstract:FABRIC is a nation-wide programmable network that will enable creation of a new, secure, flexible Internet. Running over the ESnet optical backbone with links up to 1 Tbps, it interconnects supercomputer facilities, cloud testbeds, NSF PAWR (wireless) testbeds, commercial clouds, the real Internet and powerful AI-enabled edge nodes at universities and labs. A key contribution will be FABRIC's programmability and Quality of Service (QoS).
Unlike equipment in the current Internet that simply forwards packets, nodes in FABRIC have significant computational capabilities that allow users to run computationally intensive programs throughout the network, applying application-specific routing, processing/transformation, and storing/caching of data in the network. Each node contains large amounts of compute, storage, and specialized devices such as GPUs, FPGAs, and network processors. The ability to design and deploy programs that run at any node in the network as opposed to only at the edge or in the cloud enables a completely new paradigm for developing HPC-related applications that don’t rely solely on large centers.
This panel will provide an introduction to FABRIC. We will then present four example research projects in the areas of Scientific Computing, Distributed Machine/Deep Learning, IoT and Scientific Devices, and Data
Analysis and Workflows. Panelists will describe ways to leverage FABRIC's features to significantly enhance their research. Attendees will be encouraged to think outside-the-box and brainstorm ways that FABRIC could be leveraged to advance their own research.