The path to innovation in cyberinfrastructure (CI) will require continued focus on building HPC systems and secure connections between them, in addition to the increasingly important goals of data best practices and workforce development, said Amy Friedlander, acting director of the NSF’s Office of Advanced Cybersecurity (OAC) in the final plenary session of PEARC20. On July 30, 2020, Friedlander explained how OAC’s combined focus on networking, cybersecurity and computing resources, as well as its possibly less-well-known concentration on data, software, and learning, is aimed at ensuring the nation’s CI develops with all of those imperatives accounted for.
The Practice and Experience in Advanced Research Computing (PEARC) Conference Series is a community-driven effort built on the successes of the past, with the aim to grow and be more inclusive by involving additional local, regional, national, and international CI and research computing partners spanning academia, government and industry. Sponsored by the ACM, the world’s largest educational and scientific computing society, PEARC20 took place between July 27 and 31.
This year’s theme, “Catch the Wave,” embodied the spirit of the community’s drive to stay on pace and in front of all the new waves in technology, analytics, and a globally connected and diverse workforce. Scientific discovery and innovation require a robust, innovative and resilient CI to support the critical research required to address world challenges in climate change, population, health, energy and environment.
It’s About Machines
When most people think of OAC, arguably the ecosystem of advanced computing resources funded by the office springs first to mind. Friedlander charted out the latest of these systems, and how OAC’s thinking about funding them has evolved. OAC currently funds three major classes of program: Leadership Class, Innovative Systems & Services, and Coordination Services.
OAC has funded one “Leadership Class” system, Frontera at the Texas Advanced Computing Center. As the name suggests, this system began its operational life as an extreme-scale machine near the top of what is possible with HPC, supporting research across the NSF’s directorates. As examples, Friedlander highlighted work by Alexander Tchekhovskoy at Northwestern University in astrophysics, Ivan Soltesz at Stanford in neuroscience, and Paul Morin at the University of Minnesota in analysis of polar satellite imagery that utilized Frontera.
For the first time in 2019, OAC funded two types of system within the Innovative Systems & Services class: “Capacity Systems” and “Testbed Systems.” The latter are smaller, experimental systems, the former are ones intended to go into operation and to support major work by researchers. The smaller testbed systems are prototypes to provide an on-ramp for new capabilities in OAC’s operational systems. Examples of testbeds include Ookami at Stony Brook University, announced at PEARC19 last year; and Neocortex at the Pittsburgh Supercomputing Center (PSC) and Voyager at the San Diego Supercomputer Center (SDSC), announced at PEARC20.
The Capacity Systems, meant for production computation but at a smaller scale than the Leadership systems, might spread NSF funding over more systems at more centers. The aim is to expand the opportunities and access for a broad range of supercomputing projects, enabled by the fact that new technology allows smaller systems to provide performance that previously might have been associated with larger machines. Examples include Bridges-2 at PSC and Expanse at SDSC, announced at PEARC19; and Anvil at Purdue University, Delta at the National Center for Supercomputing Applications, and Jetstream2 at Indiana University, and, announced this year.
“The notion we had was to spur innovation in two ways,” Friedlander said. “One of them was to do more, smaller systems. Given the power that is now possible, rather than building another, say, $50-million or $30-million machine, could we break these into a set of smaller systems and have them more widely distributed?” As a second prod to innovation, “we had the opportunity to have a wider range of architectures and therefore [more] ability to…innovate.”
It’s About Connections
Another top OAC funding priority is expanding and integrating these machines, other computational systems, and researchers by supporting networking and cybersecurity. OAC has enabled connectivity at virtually every scale, backing global networking via the NSF International Research Network Connections Program (IRNC); national network infrastructure in partnership with Internet2 and ESNet; and campus up to state and regional connectivity via Campus Cyberinfrastructure (CC*). Hand in hand with such cybercommunication is security, which OAC supports via the Cybersecurity Innovation for Cyberinfrastructure Innovation (CICI) program.
Friedlander called out CC* and CICI for particular focus. For the former, she cited the strong campus-level partnerships that enable learning institutions to upgrade their campus networks and external connectivity to the national research and engineering fabric. Speaking of the latter, CICI’s operational cybersecurity focus fills a gap in Secure and Trustworthy Cyberspace (SaTC), the NSF’s current, $75-million flagship security program, by funding Transition to Practice (TTP) within SaTC.
In addition, the 2020 fiscal year saw an innovation track with linkages to FABRIC and the Open Science Grid (OSG), and six additional funded projects, through Exploring Clouds for Acceleration of Science (E-CAS), which will leverage commercial cloud capabilities to perform scientific research while developing technologies needed.
It’s About Data and Software
Less well known, perhaps, among the OAC’s portfolio are the programs that fund work in data and software. Friedlander reminded the audience that the Cyberinfrastructure for Sustained Scientific Innovation (CSSI) program now integrates the Data Infrastructure Building Blocks (DIBBs) and Software Infrastructure for Sustained Innovation solicitations to support scientific innovation and discovery through improved data and software CI, following and helping to disseminate best practices in data management and software development.
“If you will, it’s another experiment by integrating these two programs under a single umbrella,” to fund operations that are flexible and responsive to the needs of the research community, Friedlander said. She’s particularly proud of how nimbly individual CSSI projects pivoted to support COVID-19 research. “My advice to all of you is, ‘Watch this space.’”
Friedlander gave specific examples of data/software advances enabled by OAC funding. GeoCODES, the Geoscience Cyberinfrastructure for Open Discovery in the Earth Sciences, a project supported by the NSF Geosciences directorate and OAC, links a number of pilot NSF data facilities with web-based metadata organizations; the computing ecosystem of NSF’s XSEDE CI program; commercial cloud services; and the most popular cloud computing tools, such as HTML5, Jupyter, R, and MATLAB. Harnessing the Data Revolution (HDR) connects foundations, systems, and CI; data intensive science and engineering; and education and workforce elements to support research in a variety of fields in the life, environmental, materials, and earth science domains. SAGE explores new techniques for applying machine learning algorithms to data from such intelligent sensors and then build reusable software that can run programs within the embedded computer and transmit the results over the network to central computer servers to support a resilient, multimessenger network to detect gradual changes in domains such as early tsunami prediction.
Mostly, Though, It’s About People
Another thrust of OAC that may not be widely known is its increasing focus on developing the sophisticated workforce necessary to maintain technical progress, Friedlander said. The office funds training at multiple levels, supporting cyberscientists, professional staff, and of course domain-expert scientists focused on their fields rather than HPC. Examples included the NSF-wide Faculty Early Career Development Program, the CISE Research Initiation Initiative, CyberTraining-based Workforce Development for Advanced Cyberinfrastructure, the CISE-wide OAC Core Research Program, and Computing Innovation Fellows.
“All of this is always about people,” she added. “It always comes back to people and we’ve known it.”
The OAC Core Research Program is a solicitation meant to foster multidisciplinary, translational research in all aspects of CI. Research areas include architecture and middleware for extreme-scale systems, scalable algorithms and applications, and an ecosystem of advanced CI. OAC’s Data-Intensive Discovery Pathways, Friedlander explained, provides the “missing middle” that links the wealth of new data sources to scientific exploration by connecting computing, data, networking, software, and above all people.
Building Infrastructure Is Slow…But the Science Is Amazing
Globally, Friedlander said, OAC is focusing on building CI that moves discovery forward securely, without becoming an impediment to exploration and open communication.
“There are matters where the policy is about the technology, and there are matters where the technology can help us operationalize solutions or point to solutions,” she said. “And then there are the third category of matters that affect how we build the technology but are far broader than the technology itself.” An example of the last being the very big question of ensuring the trustworthiness of science, which spans from best practice in data handling to larger issues of trust between scientists and the larger society, as well as gender and equity questions.
“I’ve made the case…that building infrastructure systems, despite the pace of technological change, is actually fairly slow,” she said. “That isn’t to say that we’re not calling for innovation or that all we want to do is incremental development. But it is also true, if you look back at the major developments, that they are cumulative.”