UVM completes warp-speed upgrade to DeepGreen supercomputer

-A A +A

UVM completes warp-speed upgrade to DeepGreen supercomputer

Thu, 08/01/2019 - 11:03am -- tim

UVM president Suresh Garimella, left, speaks with Mike Austin, the university's director of systems architecture, inside the Vermont Advanced Computing Core, UVM’s supercomputer. The DeepGreen cluster, which increased the supercomputer’s speed by a factor 200, is the vertical green array to the right. It is also among the fastest 100 supercomputers at any US university. VBM photo

Brandon Arcari, Vermont Business Magazine The University of Vermont announced the operational launch of its new supercomputer upgrade, dubbed “DeepGreen.” The new computing cluster is intended to massively expand UVM’s available computational power and speed by a factor of over 200, according to a press release.

Part of the Vermont Advanced Computing Core, the new cluster is 50 times more powerful than the other seven rows of computers lodged in the same room, combined.

Using a $1 million grant from the National Science Foundation, the university added a cluster of 80 high performance graphics processing units, or GPUs, containing 460,000 processing cores to the VACC, housed in UVM's primary datacenter at Technology Park in South Burlington, over the winter and spring.

The new computing cluster is now operational. It can achieve a speed of over 1 petaflop, or one thousand million million computations per second, the equivalent of 20,000 laptop computers working in tandem. 

Officials also cited the compactness and cost efficiency of the system, saying that a more traditionally designed supercomputer would have cost 15 times more, and been 24 times larger.

The system was also put together with plenty of room for expansion and thermal overhead, so the cooling needed when the supercomputer is in use is only about half of the total that the system can provide. The system has enough space to put three more server racks worth of compute power in case of a future expansion.

“This is a massive upgrade,” said Richard Galbraith, UVM’s vice president for research in a press release, “and a necessary one. In this age of big data, having a facility like this is absolutely essential for our faculty to stay at the cutting edge of their disciplines.” 

The computer will also be available for use for other colleges in Vermont and for select businesses, under certain circumstances, according to Galbraith.

“It important to realize that this will be available to faculty members from all colleges, from all departments, and it will also be something we want to increasingly make available to students,” he said. “What we have to do is pool resources and have a shared facility so that everybody in all departments and all colleges can use the equipment that is there.”

The upgrade positions the university well, said Adrian Del Maestro, associate professor of physics, director of the VACC and principal investigator on the NSF grant, in a press release.

“We think that this really a game-changer in terms of an environmentally friendly supercomputer,” he said. “It's one of the fastest supercomputers in New England, and one of the 100 fastest academic supercomputers in the country."

The computer can reproduce the conventional processing power of around 1,000 typical compute servers, while being about 95% more efficient, he said.

“We don’t have the problem that other clusters, similar clusters have, called data starvation, so we can process the data as fast as we can feed it into the GPU,” Del Maestro said. “Some of it was so state-of-the-art that we bought components that didn’t even have all the software written yet to make them work.”

According to Del Maestro, the increase in processing speed, combined with hyper-fast network connections within the cluster, will enable faculty to take on new types of research projects they did not have the computational power to explore in the past.

“It’s as important as brick and mortar laboratories,” he said. “Yolanda Chen in Plant and Soil Science can massively speed up the genome re-sequencing of the Colorado Potato Beetle to better combat emerging threats to our food supply in a changing climate.”

A visual presentation at the unveiling showed how the processing speed can read street signs and log that data just by having a camera mounted on a car.

The upgraded VACC will also be a great asset in undergraduate education, said Safwan Wshah, an assistant professor in UVM’s Computer Science department who teaches machine learning and deep learning to about 80 students a year.

“Deep Green will enable them to take on more and bigger projects and put them at a distinct advantage as they enter a job market that is tightly focused on artificial intelligence,” he said. “They are very excited to be entering this new age of discovery.”