Yale’s capacity to conduct high-performance computing and artificial intelligence-related research continues to expand, thanks to a pair of recent moves aimed at advancing the university’s computing infrastructure.
The university announced recently that it will be a partner in a large-scale, regional AI hub in Holyoke, Massachusetts, set to open later this year.
Modern science requires increasingly sophisticated tools for discovery.
Last year, Yale also joined an intercollegiate center known as the Massachusetts Green High Performance Computing Center (MGHPCC) consortium, a nonprofit, state-of-the-art data center dedicated to supporting computationally intensive research, also located in Holyoke.
Yale’s inclusion in this consortium — where it is an equal partner with Boston University, Harvard, MIT, Northeastern University, and the University of Massachusetts — marks a key milestone in its research infrastructure development, university leaders say.
At MGHPCC, members collaborate on research, pool knowledge, reduce redundancies in research computing operations, and benefit from a more sustainable source of energy for research computing infrastructure. The partnership also provides a gateway to addressing challenges in computing at a scale that are too great for one academic institution to tackle alone.
“Modern science requires increasingly sophisticated tools for discovery,” said Michael C. Crair, the university’s vice provost for research. “These tools, which make inspiration and discovery possible, often require expensive infrastructure and the expertise to run and maintain it.”
MGHPCC, located in Holyoke, Massachusetts, is an intercollegiate high-performance computing facility. It is a joint venture of Yale, Boston University, Harvard, MIT, Northeastern, and the University of Massachusetts system.
MGHPCC is the first university research data center to achieve LEED Platinum certification from the Green Building Council’s Leadership in Energy and Environmental Design program. The facility, which uses hydroelectric power, has energy saving features that include state-of-the-art temperature and monitoring controls, chilling equipment that requires less time in use than typical systems, and plumbing systems that need less water.
Organizationally, the facility is designed to allow partner institutions to share space, power usage, and cooling costs, while maintaining their own autonomy for planning and conducting research.
“You can think of it as being like a condominium complex,” said John Barden, Yale’s vice president for information technology. “MGHPCC operates as the condo association. But the individual computers we have there are owned and operated by each individual institution to serve their faculties’ unique needs.”
By this June, Yale expects to have a second high-performance computing cluster operational at MGHPCC. That cluster will have an emphasis on secure biomedical data, including genomic information that requires compliance with more extensive regulatory and review protocols.
I view this as a positive development for the growing number of researchers across all fields who want to make more use of computational resources.
The new AI hub, also located at the Holyoke site, will differ in several ways, most notably in that it will be a shared computing system. That system, Barden said, will be more capable and much larger than any single institution could build. This, he said, will provide a unique asset to strengthen Yale’s ability to support its scientific objectives.
The AI hub is projected to have between 1,500 and 2,000 graphics processing units (GPUs), operating much like a national laboratory in which partnering institutions will be able to access the cluster individually or as collaborators. Yale’s partners in the Massachusetts AI hub will include the MGHPCC member institutions, as well as stakeholders from Massachusetts state government and industry.
“We went to our faculty, who had advised on the original AI technology task force infrastructure recommendations, and asked if they would support this sort of framework, and the answer was a resounding ‘yes,’” Barden said.
Rajit Manohar, the John C. Malone Professor of Electrical & Computer Engineering and Computer Science at the Yale School of Engineering and Applied Science, has been using Yale’s new computing resources at MGHPCC for several months in his computer chip design research.
“I view this as a positive development for the growing number of researchers across all fields who want to make more use of computational resources,” said Manohar, who is faculty chair of an advisory committee looking at research computing issues on campus.
Manohar said it is “critical” that Yale expand its computing infrastructure — particularly in ways that are environmentally conscious and offer a balanced approach to computational resources for researchers in the physical and applied sciences, and the humanities.
“This gives us more bang for the buck,” Manohar said.
The additional technical infrastructure is part of a $150 million investment over five years, which the university announced last August, to support faculty, students, and staff in engaging with AI. The commitment responds to the report of the Yale Task Force on Artificial Intelligence. For the report, an 18-member group of faculty and campus leaders engaged with dean-led faculty panels and university experts in education, collections, clinical practice, and operations to review AI activity already underway and develop a vision for Yale’s leadership in the future.
“Yale has long been at the forefront of AI development and research, and our leadership continues to be necessary as this technology evolves and endures,” Yale Provost Scott Strobel wrote in a message to the Yale community, providing details of the efforts. “To fulfill the university’s mission to improve the world and prepare the next generation of society’s great leaders and thinkers, we must explore, advance, and harness AI for its benefits while providing ethical, legal, and social frameworks to address the challenges it poses.”
