No Coding Skills Needed, Supercomputer Builds ‘Bridges’ To Non-Traditional Users

Jan 24, 2017

PSC's Nick Nystrom points out details of the Bridges Supercomputer to CMU computer science professor Tuomas Sandholm.
Credit Michael Henninger / CMU

Supercomputers help people crunch big data in a number of fields, from research to weather forecasting, but not everyone has access to one or the technical savvy to make it work. Though, a new supercomputer could offer more access.

Pittsburgh Supercomputing Center just launched the second phase of its supercomputing system, called “Bridges,” last week. It’s among the largest supercomputers in the U.S., but according to Pittsburgh Supercomputing Senior Director of Research Nick Nystrom, it is probably the most democratic one available.

“Bridges is intended to make supercomputing available to everyone,” Nystrom said.

In fact, one scientific director suggested calling it “My big laptop.”

“That’s the vision we had,” Nystrom said.

Most supercomputers require users to write their own code in order to run the machine. That means researchers usually need to have at least one computer scientist on their team just to operate the system. Bridges uses portals that will crunch numbers for just about anyone with a big set of data. No technical language required, all the user needs is an internet connection.

“People can interact with them just through a web browser using language and paradigms that they are already familiar with,” Nystrom said. “It lets them use Bridges without knowing they’re using a supercomputer. Basically just interacting with the web, even from a tablet.”

Getting time on the supercomputer does take a bit more than just pulling up a website. Users' proposals must still be reviewed and it could take hours or days waiting in line to get the needed time on the servers. However, CMU Alumni University Professor Clark Glymour, who’s in his seventies, joked that it’s even easy enough for him to use.

And having access to such computing power will be indispensable going forward, he said.

“It is almost impossible to make a serious contribution that will affect methodology or inference without computational skills,” Glymour said.

Glymour’s research involves mapping and tracking electrical activity in the brain based on massive amounts of imaging data.

“The people at the supercomputing center are just fabulously helpful,” he said. “They really want their tool used.”

Nystrom said the time saved with a supercomputer is crucial, likening them to time machines.

“Because they let people accomplish problems today that, if they were to wait for their local resources to become affordable and to catch up, they would catch up,” Nystron said. “But it would take five or 10 or more years to do so and a lot of research can’t wait for that.”

Researchers in virtually every field are using big data and the PCS built Bridges to not only crunch that data, but also to store it, which is unusual for a supercomputer.

One of the largest users of Bridges is the University of Pittsburgh Cancer Institute’s genome repository, which is looking for exploitable patterns in cancer DNA. That project has been allocated 2.5 petabytes of memory, or 2-and-a-half million gigs of cancer gene sequencing.

The first phase of the Bridges supercomputer went into operation in February of 2016, the second phase will start operating Friday.

It’s the 18th supercomputer run by the Pittsburgh Supercomputer Center since it opened in 1986. Generally, as soon as a supercomputer is up and running, the managers start planning for the next generation.

“We will absolutely keep going, providing national scale computing resources. I think we have a lot of insights into what the research community needs,” Nystrom said. “And that’s key because we don’t build large systems just to have a system, we build it to solve problems.”

Nystrom said it is very possible that the next system will focus on enabling machine learning and artificial intelligence research.

In this week's Tech Headlines: 

  • team of student researchers from the University of Pittsburgh has captured a gold medal at the International Genetically Engineered Machine Foundation’s annual competition. The award came for the group’s work in developing an inexpensive way to test lead toxicity of drinking water in people’s homes. The project uses “bacterial cell extract” and a “DNA genetic circuit” to detect high levels of lead in water. Nearly 300 teams from several countries took part in the competition.
  • A computer virus that destroyed systems of the Saudi Arabian state-run oil company nearly five years ago has returned to the kingdom, with at least one major petrochemical company apparently affected by its spread. In 2012, it was suspected that the virus came from Iran because it came soon after the Stuxnet cyberattack targeting Tehran's contested nuclear enrichment program. A report this week by Saudi state-run television included comments suggesting that 15 government agencies and private institutions had been hit by the Shamoon virus. Computers affected reportedly had their hard drives erased and displayed a photograph of the body of 3-year-old Syrian boy who drowned fleeing his country's civil war. Iran denied being responsible for the 2012 outbreak. Tehran had no immediate comment on the new attack.

The Associated Press contributed to this report.