USING SUPERCOMPUTERS TO ADVANCE RESEARCH
One of the latest universities to boost its computing capabilities is Stony Brook University in New York, which deployed a high-performance computing (HPC) system built with new components developed last year by tech companies Hewlett Packard Enterprise (HPE) and Intel. According to a June announcement, the new HPC system was made possible through about $485,000 in grant funding from the NSF, as well as additional funds from the New York State Department of Economic Development, and will be used mostly for research projects.
Robert Harrison, director of Stony Brook University’s Institute for Advanced Computational Science (IACS), said the recent upgrades built upon over a decade of efforts to ramp up the university’s supercomputing capacity for research. Among other projects, he said, the institute installed a supercomputer called Ookami in 2020. He added that the newly upgraded HPC system will work as part of the university’s SeaWulf computational cluster, which uses advanced components from companies like Penguin, Intel, Nvidia and Mellanox.
“Seawulf is open to all researchers and students basically for free on campus. Over time, it’s grown to be where it is now, which is about a two-petaflops system,” Harrison said, referring to floating-point operations per second, a measure of computer performance. “The memory is somewhere between five and eight times faster than the standard memory that you get on most servers.”
He noted that about 40 academic departments are using the university’s supercomputing capabilities for research in fields such as computational science, astrophysics, climate science, social sciences and linguistics, among other research areas.
“Everyone’s computing these days,” he said. “Lots of more recent applications are of course data oriented, so we’ve got people analyzing Twitter feeds and data, Facebook and elsewhere, looking and trying to understand social trends, either nationally or internationally. … We’ve got all the simulation work going on in computational chemistry, drug discovery and things like that, and then you’ve also got people of course doing machine learning, and that’s a growing area [of interest].”
Noting similar benefits in terms of accelerating research, Montana State University also recently used $500,000 awarded by the NSF to upgrade its central processing units, as well as new graphics processing units from NVIDIA that are “10 times more powerful” than the university’s old GPUs. According to an announcement in January, the funding was awarded to the university’s Research Cyberinfrastructure Core Facility, which manages advanced systems for research in quantum science, large data analysis, machine learning and artificial intelligence technologies.
Coltran Hophan-Nichols, the facility’s director, said these upgrades “roughly double” the university’s overall research computing capacity, particularly for researchers in fields such as quantum science, precision agriculture and, of course, artificial intelligence and machine learning.
“We certainly expect to see the trajectory in research activity and the corresponding computational demands of that research continue growing, so we expect that projects like this will continue [with more upgrades] going forward,” he said. “I can’t speak specifically to what the next one might be, but we’ve been steadily increasing our capacity as our institution’s research activity has increased.”
INTEGRATING SUPERCOMPUTERS INTO LESSONS
Hophan-Nichols said that in addition to giving students and researchers access to the most powerful high-performance computing capabilities available in the state, Montana State’s supercomputer system can also be used for instructional purposes.
“While the main focus of this has been on research [uses], one of the other things that we’ve been seeing and that we are addressing with this recent grant is a tangible, expanded use of HPC in the classroom. Over the last few years, we’ve had a significant increase in the number of courses and the number of students in those courses that are making use of HPC in the classroom, either to learn the technology itself or as a tool for their field of research,” he said. “We’re seeing that in places like computational chemistry, geospatial information systems. … Enabling these tools for real-world experiences, for both graduate and undergraduate students in their courses, has been another important area.”
The University of Florida is using its recently upgraded HiPerGator supercomputer system for similar instructional applications, according to a January news release that said more than 300 instructors from Florida’s 12 public universities have used the supercomputer as a means of introducing students to AI and machine learning applications. The release noted that the supercomputer, reportedly “one of the fastest AI supercomputers in academia,” has played an integral role in ongoing efforts to make the University of Florida into the country’s first “AI University” by integrating AI across all areas of university research, teaching and curricula.
“The powers of HiPerGator are twofold: We are pushing the limits of modern research while also training the next generation to build on the groundbreaking work that’s being done now,” Erik Deumens, the university’s research computing senior director, said in a public statement. “In the coming years, we’re going to be solving problems and breaking barriers at a faster rate than ever before.”
Harrison, from Stony Brook University, said he expects to see more universities increase investments in developing and launching their own supercomputing systems in the years to come, especially as AI becomes a focal point of research and education initiatives.
“Having stuff on-prem, you can operate it very cost effectively, and you’ve got a lot of flexibility in how you use it,” he said. “You’ll see continued and growing investment in this stuff, not just at Stony Brook, but everywhere.”
