Computer recycling system has educational, cost benefits

Andy Howard & Fengping Hu
Download photo
caption below

"This program is using the state's funds very effectively," said Jeff Evans, an assistant professor of electrical and computer engineering technology. "That's the beauty of it — the money for the computers has already been spent, and what we are doing is basically using the machines until they break to the point where we can no longer repair them."

This system, which has been in use for about a year, was developed in collaboration with Purdue's Rosen Center for Advanced Computing. The process allows computers that are about six or seven years old — too obsolete for faculty and students to use for complex scientific research — to be collected for electrical and computer engineering technology students to practice software and hardware installation, benchmarking, troubleshooting and maintenance in the emerging area of high-performance computing.

As a result of the students' work, the high-performance computer the students continually work on benefits researchers across disciplines who need computing power to perform large tasks.

Evans said this system of using recycled computers for a high-performance computing teaching lab is the first of its kind at Purdue and that there are very few such systems of this scale in place around the country.

In addition to saving money, students also benefit. The undergraduate students involved are part of Evans' "Electrical and Computer Engineering Technology 325" class, which has been offered for about a year on the West Lafayette campus and was expanded this fall to the College of Technology's statewide location in South Bend. The students there are able to link to the same supercomputing system used by students in West Lafayette.

A portion of the class exposes students to the area of high-performance computing, which Evans describes as the process of linking standard PCs together to form a Linux cluster (a group of linked computers that work together to operate as a single computer), creating a supercomputer that is able to perform complex tasks better and much more efficiently than in the past.

"These formerly unrelated computers are linked together to act and be maintained as one," Evans said. "The result is that tasks and research that would have been difficult or impossible in the past are now possible. High-performance computing allows a task that would have taken years in the past to be performed in just hours."

This computer recycling system is an extension of a similar effort that has been going on for about four years at Information Technology at Purdue's Rosen Center, which supports large-scale scientific computing and storage for the Purdue research community.

Before the recycling system was fully in place, the process worked slightly differently. Purdue's Teaching and Learning Technologies would purchase new computers to be placed in student computer labs that are operated by Information Technology at Purdue (ITaP), said Bill Whitson, ITaP's manager of research computing. When those computers became old or obsolete, they were given to the Rosen Center to be housed, linked together and used as a high-performance computing resource.

Whitson said the idea to expand the effort was developed after Evans arrived at Purdue in 2003. Evans is involved in research with distributed and high-performance computer systems and was in need of computers, and the Rosen Center was experiencing difficulties finding room and cooling capacity to accommodate the computers it inherited.

"The program was almost a match made in heaven," Whitson said. "We were running out of space, and Jeff was eager to take these machines off our hands and put them to good use."

Evans' class is an offshoot of his research, which focuses on factors that create performance degradation in distributed computing systems and how to evaluate and improve performance. His recent research has included the impact of network performance degradation on parallel application run time.

"For instance, we are developing ways for researchers to easily quantify the sensitivity of their applications. This metric then has the potential to be used by other parts of the system to manage overall system performance," Evans said.

"Researchers know their subject areas and what they want to accomplish very well, but often rely on simple and optimistic models for the computing part. What my research involves is trying to manage complex interactions within the computing system itself to make the researcher's job easier."

The class gives undergraduates an opportunity to study areas that will be of increasing importance as this technology expands.

"Supercomputing is a growing field, and training in this area — especially for undergraduates — is not offered in very many places," Evans said.

Some of the jobs that electrical and computer engineering technology students can pursue include circuit design, product development and testing, and computer software and hardware sales and maintenance.

Students in Evans' class not only learn about the theory of how to link the computers, but also the practice, which involves modifying software, hardware, troubleshooting and measuring performance.

He said these students will have a major advantage when they enter the work force because high-performance computing is becoming much more commonplace in corporations.

"There are great employment opportunities in this area, and organizations that support high-performance computing are having difficulty finding qualified system administrators," Whitson said. "We think this program could help turn that around."

Whitson said the Rosen Center currently operates about 400 recycled machines, down from a maximum of 1,008 in 2004 due to data center space and cooling constraints. Evans has nearly 200 machines in his Adaptive Computing Systems Lab in the Michael Golden Engineering Labs.

"It's been a very successful concept," Whitson said. "The reactions from other universities have been, 'Why didn't we think of that?' Now, several other universities are creating similar systems to Purdue's."

Evans said high-performance computing is being applied in a wide variety of areas, such as DNA sequencing, financial modeling, fluid dynamics simulation, protein analysis, astrophysics, and climate and weather modeling.

A course for graduate students, which would focus more on how to build a supercomputer, currently is being developed. Evans said there also are plans to expand the undergraduate course to additional College of Technology statewide locations.

Writer: Kim Medaris, (765) 494-6998, kmedaris@purdue.edu

Sources: Jeff Evans, (765) 494-7725, jje@purdue.edu

Bill Whitson, (765) 496-8227, whitson@purdue.edu

Purdue News Service: (765) 494-2096; purduenews@purdue.edu

A publication-quality photo is available at https://www.purdue.edu/uns/images/+2006/evans-computertech.jpg

To the News Service home page