University of Alaska Fairbanks

10/03/2024 | Press release | Distributed by Public on 10/03/2024 11:05

UAF's high-performance computing resources just got a lot smarter

UAF's high-performance computing resources just got a lot smarter

Rod Boyce
907-474-7185
Oct. 3, 2024

Photo by LJ Evans
Kevin Galloway, left, and Mike Steinhauer of the UAF Geophysical Institute's Research Computing Systems unit attach cabling for the final GPU node to be installed as part of a major upgrade to the high-performance computing cluster Chinook.

University of Alaska Fairbanks staff upgraded the university's most powerful high-performance computer this summer to significantly increase its capacity and usefulness to researchers and students.

"These upgrades are specifically designed to make it possible for researchers and students to use the same artificial intelligence and machine learning tools that are behind applications like ChatGPT, Google Gemini and Hugging Face," UAF Geophysical Institute Research Computing Services Manager Liam Forbes said. "These applications and services have become very visible the last couple years."

Technicians installed six new high-end compute nodes with 48 graphics processing units, or GPUs, in the university's Chinook computer system. The work was part of the Cyberinfrastructure for Big Data Research project initiated by faculty and staff in UAF's Computer Science Department and the Geophysical Institute. Personnel from the Geophysical Institute Research Computing Services unit, the University of Alaska Office of Information Technology and the UAF Computer Science Department installed the upgrades.

Funding of almost a million dollars for the upgrades came from a National Science Foundation Major Research Instrumentation grant proposal submitted by Arghya Kusum Das, faculty member in UAF's Computer Science Department, and Forbes. Parts started arriving this spring, and the final and largest new GPU compute node was installed in late August.

The project will also install 1.3 petabytes of new data storage later this fall. One petabyte equals 1 million gigabytes. Personal computers generally have 8 to 32 gigabytes of memory and 1 to 8 terabytes of data storage.

Das and Forbes wrote in their proposal that the upgrades will be "a catalytic instrument in increasing high-priority, data-driven research capacity in multiple areas of science and engineering disciplines critical to Alaska, its climate and natural resources, including its distinctive flora and fauna, and its sociocultural and economic mobility."

The upgrades also will aid collaboration between UAF and its research and education partners across the state.

Das said 26 co-authors helped write the proposal and almost 50 scientists are already queued up to use the increased computing power and storage space.

"The best part about working in Alaska, especially working in UAF, is that there are lots of good data right here that can be processed with this expanding computing power," Das said. "With this upgrade we are bringing both expanded storage and the latest computation to the same location."

Photo by LJ Evans
Liam Forbes, manager of Research Computing Systems at the UAF Geophysical Institute, points out some of the features in the final and largest GPU node to be installed as part of a major upgrade to the high-performance computing cluster Chinook.

As of the start of the semester, the first researchers and professors at both UAF and UAA were working out how to use the GPU compute nodes and to beta test the software and libraries. The Computer Science Department is creating a Data/AI Lab to help students, scientists and educators.

Das is developing a series of classes to enable students, both undergrad and grad, to use the new computing power. He is already teaching a deep-learning course where the students will use the new GPUs.

"They are very much excited about it," Das said. "I have one Ph.D. student who has already started utilizing it. We have a lot of forest fire data, and there are lots of videos. We were tracking the direction of the smoke, calculating the speed of the smoke, speed of the fire spread, and other things. There was no way possible (to do this) with a simple server or laptop."

The Geophysical Institute launched Chinook in 2016 as an energy-efficient Linux cluster purchased fromPenguin Computing Inc. Chinook is intended for use by UAF researchers and has been upgraded several times since then. These latest upgrades are some of the most significant in terms of computational power.

Chinook is named in honor of longtime GI colleague Kevin Engle's unique, strong, collaborative nature and passion for salmon and Alaska. Its initial purchase was made possible through funding from the GI and other units at UAF, including the International Arctic Research Center, Institute of Arctic Biology, the UAF Vice Chancellor for Research, and researchers who pay for and share computational resources. In 2016 and 2017, Chinook expansions were also supported with funding from the M. J. Murdock Charitable Trust.

ADDITIONAL CONTACTS: Liam Forbes, [email protected]; Arghya Das, [email protected]

NOTE TO EDITORS: Photographs and a longer version of this news release are available at the Geophysical Institute website.

075-25