Parallel Programming

Computer science students successfully networked 44 computers to create a single, extraordinarily powerful, virtual cluster computer.

By Kalie Caetano ’13

Computer science students in Professor Libby Shoop’s summer research group successfully networked 44 computers to create a single, extraordinarily powerful, virtual cluster computer. These students co-authored a paper to be presented by Professor Shoop at the Association for Computing Machinery (ACM) Conference on Computer Science Education Feb.–Mar. 3, 2012 in Raleigh, N.C.

“I personally deal better with machines,” says Devry Lin ’12, who, along with fellow computer science students, Malcolm Kane ’13, Maura Warner ’12 and Eric Biggers ’14, devoted a whole summer to dealing with machines.

“Think about it as a kitchen at a restaurant,” says Malcolm Kane ’13 (Ridgefield, Conn.). “The master computer is like the head chef. The master computer distributes tasks, and then the other computers follow those instructions.” The purpose is to pool multiple computers’ processing power in order to tackle more challenging computational tasks.

“It’s just a tool for breaking down really big, intractable problems into manageable chunks,” says Maura Warner ’12 (Greenbelt, Md.), who began working on the project two summers ago.

The project, headed by Professor Libby Shoop and in collaboration with St. Olaf Professor Dick Brown, was funded by the National Science Foundation. Over the summer the student team tussled with the project, troubleshooting and reworking until they had successfully networked 44 computers in the math labs of Olin-Rice Science Center, gaining valuable firsthand experience in building a working cluster of virtual machines with separate operating systems running inside each physical computer.

The incentive for pulling together these networks is to increase computing speed by overcoming the limitations of current processor speeds. For many years, computer processors had increased in speed at an exponential rate, but this finally tapered off once the silicon chips inside began to overheat. These physical limitations turned computer manufacturers away from focusing on increasing processors’ speed and instead led them to consider multi-core processors—including two or more processors in a machine (which is what is meant by “dual-core” on manufacturers’ labels).

“Computers will contain many processors and be networked together into clusters from now on,” says Shoop, “So our graduates must be aware of how to get the most out of these machines.”

“Everyone on the team got something different out of the experience,” says Lin (Taipei, Taiwan). “This project gave me a better grasp of how the hardware and the software of the computer interact.”

Delving into a technical hands-on project improved Kane’s programming ability. “Computer science is one of those things where the best way to get better is through practice and trying new things,” he says.

Now armed with the technical know-how to create a virtual cluster from scratch, Warner noted that “not a lot of computer science majors graduate with this kind of knowledge.”

“I thoroughly enjoyed working with this outstanding team of students,” said Shoop. “Because of their hard work, I am able to teach students at all levels how to use this system of computers. And best of all, we are sharing our experience with computer science professors throughout the country, so that they can do the same.”