High performance computers worldwide enable researchers in various fields to solve the riddles that plague the modern world. At SDSU, an upgrade to the existing high performance computer could solve more complex problems and open a wealth of new possibilities for researchers and students.

SDSU has been home to a high performance “supercomputer” since 2003, with a second cluster that was added to it in 2007. Recently, grant money from the National Science Foundation has allowed Information Technology Vice President Michael Adelaine to pursue a more advanced supercomputer on campus.

The NSF grant allowed Adelaine to upgrade the networking components of the supercomputers, including 10-gigabyte fiber optics between research buildings and the supercomputers, better cooling and air-conditioning, generators and other components. Adelaine said because of the grant, the money he would have had to use for the networking was free for use elsewhere.

“The grant isn’t providing the money for the cluster, but what it did is free the money I would have had to use to replace (the networking equipment) so that I could use it for the cluster,” Adelaine said.

Improving the supercomputer on campus would allow researchers to tackle larger and more complex algorithms and problems than what the current system could. It also allows them to scale up their algorithms on campus instead of off it.

“We can buy time on the world’s largest supercomputers but you don’t want to spend your time getting your program working. You want to spend your time getting your data and getting your answer out,” said George Hamer, assistant department head for the Department of Electrical Engineering and Computer Science. “What this will give SDSU is the capability of debugging the program to make sure it works properly and it works on the large scale.”

With the increased computing power, Adelaine said more researchers and departments will be able to work with the supercomputer. He also said professors will be able to give their students experience working with supercomputers.

“This stuff is new enough that it’s going to be the students that are going to be the generation that really brings this forward and we’ve got to begin training them and letting them have experiences so they can help us,” Adelaine said.

Updating the supercomputer system began last summer with the purchase of the housing unit for the supercomputer processors. Costing $250,000 just for the base, Adelaine said the third cluster base was originally going to be filled by researchers as they received grants to buy the blades, or processor units. However, Information Technology realized the efficiency of the supercomputer would be hampered by having different models installed over a period of time.

In order to achieve the highest efficiency, Adelaine said he hopes to populate the rest of the base within the next couple of months with homogenous blades, which are composed of six core processors. The cost to fully populate the base would be $416,000 – money freed from other expenses by the NSF grant – and would add an additional 864 core processors to the current 336.

The third cluster would increase the current speed of the supercomputer to 18 teraFLOPS (Floating-point operations per second), which means the supercomputer could perform 18 trillion mathematical equations every second. This would increase the current supercomputer’s power by more than five times. An algorithm that takes 20 minutes today will be able to run in a matter of seconds and answer broader questions, Adalaine said.

Even with these upgrades, Hamer said the supercomputer would just barely make it to the bottom of the top 500 list of supercomputers in the world.

“What this would give us is the capability of reaching the bottom of that list. You need to get up to at least that level to do the serious high performance computing,” he said.

Reaching the lower end of the top 500 list means that SDSU will become more appealing to prospective faculty, staff and students, thereby drawing in the best students and faculty.

“It’s a recruiting tool both for students and faculty. We want to attract the brightest students to SDSU and obviously we want to attract the brightest faculty. Whether it’s a student or a faculty member, if we don’t have the base level of infrastructure, they’re going to go to another university,” he said.

Hamer also noted that SDSU was “alone even at the current level” when compared to computing systems at other universities in South Dakota.

Daniel Vellek, Students’ Association senator for the College of Engineering, represents the Students’ Association on the Information Technology Committee, of which Hamer is the Chair. Being involved in the committee, he has been a part of the supercomputer deliberating process. He believes the supercomputer would be a great decision by the committee and that students in all the sciences would benefit.

“In the long term, we’re talking about chemistry and biology majors being able to run these programs and having them take maybe a C++ class so that they can do their research as well. It’s not just engineering that uses computing power, you can use computing power for any type of research,” he said.