Award Abstract # 2005506

ABSTRACT

The frontiers of science are rapidly evolving in regard to availability of data to be analyzed and the breadth and variety of analytical tools that researchers use. To effectively analyze and make sense of this ever-growing cache of information, and to make it possible to leverage new artificial intelligence tools for research, researchers need on-demand, interactive, and programmatic cyberinfrastructure (CI) delivered via the cloud. Jetstream 2 is a system that will be easy to expand and reconfigure, and capable of supporting diverse modes of on-demand access and use, The system will also revolutionize the national cyberinfrastructure (CI) ecosystem by enabling “AI for Everyone” with virtual GPU capabilities and widespread outreach through the five partners, led by Indiana University. The project promises to enable the research community to use a greater variety of computational resources and to expand its reach into student populations, drawn from a broad range of disciplines, thus contributing to building the future STEM workforce.

Jetstream 2 will be an 8 PetaFLOPS (PFLOPS) cloud computing system using next-generation AMD “Milan” CPUs and the latest NVIDIA Tensor Core GPUs with 18.5 petabytes (PB) of storage. Consisting of five computational systems, Jetstream 2’s primary system will be located at Indiana University, with four modest regional systems deployed nationwide at Arizona State University (ASU), Cornell University, the University of Hawaiʻi (UH), and the Texas Advanced Computing Center (TACC). Additional partnerships with the University of Arizona, Johns Hopkins University, and University Corporation for Atmospheric Research (UCAR) will contribute to Jetstream 2’s unparalleled usability and support for a broad range of scientific efforts.

The Jetstream team has been at the forefront of training the research community to transition from batch computing methods to adopt cloud-style usage. Jetstream 2 will continue this path and will ease the transition between academic and commercial cloud computing. Some of the advanced features include push-button virtual clusters, advanced high-availability science gateways services (including commercial cloud integration), federated authentication for JupyterHubs, bare metal and virtualization within the same system through programmable CI, support for on-demand data intensive workloads in addition to on-demand computation, high-performance software-defined storage, and advanced multi-platform orchestration capabilities.

Jetstream 2 will have far-reaching societal benefits. As enhanced educational infrastructure, it will serve more students, from traditional undergraduates to domain-science experts desiring training in computational techniques, than any other NSF-funded CI resource. These students will be better equipped to fully participate in the evolving STEM workforce. In addition to enabling new research, discovery, and innovation across many disciplines, Jetstream 2 will advance the national CI ecosystem and extend the broader impacts of existing NSF investments. Jetstream 2’s “Core Services” will demonstrate a practical model of distributed cloud computing that will give academic institutions an incentive to invest their own funds in new advanced CI facilities. Although modest in scale, these facilities will represent the state of the art in reconfigurable computing. The implementation of Jetstream 2 will also demonstrate that colleges and universities can invest sustainable amounts of their own funds in highly-effective, flexible CI resources that generate a significant return on investment. In sum, Jetstream 2 will transform the national CI landscape and greatly benefit the nation.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Please report errors in award information by writing to: awardsearch@nsf.gov.