UC Merced is rapidly gaining a strong reputation for research and scientific computing across many disciplines and a major expansion of its computing infrastructure is about to cement the campus’ status as a research computing hub.
The National Science Foundation (NSF) is supplying $700,000 and the UC Merced Office of Research and Economic Development and the Cyber Infrastructure and Research Computing unit in the Office of Information Technology are adding the equivalent of $300,000 so that, over the next three years, the campus can purchase both hardware and software for a new shared high-performance computing cluster called “Pinnacles.” This is the third NSF award to chemistry Professor Hrant Hratchian and applied math Professor Suzanne Sindi, interim co-directors of Cyberinfrastructure and Research Technologies (CIRT), to support large-scale research computing infrastructure at UC Merced.
“This will more than double our computing capacity over the next three to five years,” Hratchian said.
In 2015, NSF awarded the campus about $600,000 to build the Multi-Environment Research Computer for Exploration and Discovery, or MERCED — a cluster of individual computer nodes, each featuring different kinds of computing architecture.
The supercomputer is expanded and updated as researchers purchase nodes that are specific to their work, expanding capacity without buying a whole new cluster. Right now, MERCED has more than 3,000 processing units, which means researchers can run more than 7.2 trillion calculations per second — that’s more than 1,500 times the number of calculations per second per core than a typical laptop can perform. More than 370 terabytes of data can be stored on MERCED — about 370 times more data than can be stored on a typical laptop.
The campus supercomputing clusters have hundreds of users.
“It’s actually a small expense for the number of people it supports,” Sindi said. “We have a huge impact on the campus.”
Computing is playing an increasingly central role in knowledge discovery and scientific advancement and is one of the few research activities that has been able to continue during the COVID-19 pandemic.
“The Office of Information Technology appreciates the opportunity to partner with faculty on advancing their research interests as technology is both an enabler for computational research outcomes as well as an instrument for discovery,” Associate Vice Chancellor and Chief Information Officer Ann Kovalchick said. “In addition, learning about faculty research makes our work meaningful and intellectually engaging.”
Sindi, Hratchian and mechanical engineering Professor Ashlie Martini are the co-leads on the grant.
High Performance Computing Manager Sarvani Chadalapaka said the expansion means more access for students at all levels and more student training opportunities, as well as serving as the centerpiece for high-impact work in three research areas: modeling and simulation of complex systems; data-enabled science; and numerical optimization.
“While the fundamental scientific avenues pursued in each of these pillars traditionally do not overlap, a particular strength of the project's approach is a common emphasis on methodology and high-performance computing,” Chadalapaka said. “The Pinnacles cluster will facilitate knowledge discovery by supporting researchers in their individual research areas and enhancing interdisciplinary collaborations to strengthen advancements at the boundaries of traditional disciplines. This will support and enhance scientific computing in all three schools at UC Merced.”
The CIRT office trains student researchers on supercomputer use, software optimization and many other topics. It offers customized training for researchers at all levels of supercomputer use across campus, as well as workshops on software and data carpentry that focuses on using Unix, Git, Python and R, as well as data-management planning in partnership with the UC Merced Library. The office also provides one-on-one consultations with researchers to help design projects, improve workflow and develop data management plans for successful grant applications.
One advantage of using a supercomputing cluster is that computational research work can be done from almost anywhere with a reliable network connection.
Providing that ability is more important than ever right now, Chadalapaka said.
“We’re very proud of being able to provide resources and tools that make it easier for researchers to embrace asynchronous learning,” she said.