Upgrades to UVM supercomputer 'Deep Green' propel research
The University of Vermont has completed a million dollar upgrade to its supercomputer, making it one of the fastest academic supercomputers in the country.
To some it may look like something out of Star Trek. "This is one of the coolest things I think I have ever seen in my life," said UVM grad student Dillon McCarthy. He's one of the few students who got early access to the school's upgraded super computer, Deep Green.
"It's already done stuff that we're hopefully going to publish really soon, so within a month we've already gotten results out that are worthwhile and scientifically relevant," McCarthy said.
Deep Green is a cluster of 80 high performing graphics processing units or GPU's. One hundred laptops can fit into just one of these small GPU's. When they are all working together, Deep Green can reach speeds of over one petaflop. "That's a one with many, many zeroes behind it," McCarthy said.
That's the same as 20,000 laptops working at the same time. It can take in and process more data faster to allow researchers to do more.
"In science, you have an idea, you want to test it then if it didn't work. You want to change something, and so there's this feedback loop that we do as scientists that let's us get to the solution of a problem and if you have an idea and you have to wait two months, by then you're onto a totally different thing," said Adrian Del Maestro, an associate professor and director of the Vermont Advanced Computing Core.
One example of how UVM research benefits from Deep Green includes a project to build a real-time map of highway signs looking at quality and effectiveness. Deep Green allows them to read signs in an instant and collect data, which would usually take more man-hours and computing time. It can also help process research on artificial intelligence and new medications. "Kind of look at things that we want to see, like how proteins interact with drugs if we want to design new drugs," McCarthy said.
Del Maestro says the National Science Foundation funded investment in cyber infrastructure is essential to the university for research purposes, and he hopes the cutting edge technology will be useful for five years or longer. "Other institutions are interested in building other supercomputers and they're saying, 'How did you do it, can we learn from which mistakes you made to make this thing?' We said, we didn't make any mistakes. we built the perfect supercomputer for Vermont," he said.
That is, until even newer technology is developed.