Computers UNB researcher working on Java Virtual Machine efficiency

B1

Paola Loriggio
Telegraph-Journal

Stumble Upon

del.icio.us

Digg

Facebook

Print

Email

Speak Up

University of New Brunswick researcher Kenneth Kent plans to make computer systems - from the top online shops to government servers - run faster.

Enlarge Photo

Keith Minchin/For the Telegraph-Journal

Kenneth Kent, a computer science professor at UNB, has spent years improving Java Virtual Machine.

The computer science professor has spent years improving Java Virtual Machine, a set of computer software programs and data structures that allow Java applications to run on various operating systems. Java is a popular programming language with widespread commercial use.

Now he's preparing to take the helm of a new centre at the university's Fredericton campus that will work to speed up Java Virtual Machine on large, complex computer systems.

The centre, a partnership with IBM, will have some 20 staff and employ about as many graduate students from UNB's Faculty of Computer Science, according to a statement by the university. The project will receive roughly $3 million from the Atlantic Innovation Fund over four years. Total costs should reach $5.1 million.

Eventually, the centre "will have a larger team working on Java technologies than IBM has in its commercial shop in Ottawa," Kent said in a phone interview from his campus office.

The project could bring short- and long-term jobs to Fredericton, boost the university's reputation as a leader in technological research and reduce the province's brain drain in computer science, Kent said.

He wants to show students they have options in the province, and hopes IBM will set up a permanent facility in Fredericton if the project proves successful. "We produce good workers, but they're always running away to Ontario," he said.

IBM spokeswoman Carrie Bendzsa said the company "works with specific professors on specific projects," but wouldn't say whether it would set up shop in New Brunswick after the project is over.

The centre is "a recognition of the high calibre of the staff and research taking place at UNB," Bendzsa wrote in an email. "Some of our new graduates working on cutting-edge projects like the Java Virtual Machine come from UNB."

For students, the centre is a foot in the door of a top company as it races to enhance efficiency on multi-core computer systems, Kent said.

Computer companies can't make processors - also called "cores-" much faster, so they boost performance by linking several processors, he said. Some gaming systems boast 32 processors, he said.

As multi-core systems gain in power, they often lose in efficiency, the computer science professor added. Most problems concern memory and communication, he said.

"It's like going from one employee to 32," he explains. "You have two challenges: where to put the information so everyone can get it, and how to make it so that all the workers are talking to each other."

The centre will work to identify bottlenecks and low-performance areas while running Java Virtual Machine technologies on multi-core systems, Kent said. The team will then design algorithms to fix these problems.

He expects some progress within a year, but the bulk should occur in the third and fourth years. The software tools and techniques created by Kent's team could hit the market quickly, thanks to the direct link with IBM, he said.

"Practically everything would go faster," he said of the impact on consumers, noting Service New Brunswick's back-end servers run on Java. Online stores such as Amazon also include Java applications.

Kent came late to the world of computing. While at Memorial University, he studied history and math, with the ultimate goal of becoming a teacher, like his parents.

He took a mandatory computer science class as part of his math coursework. "I liked it, and I did well, so I thought, 'I should do another one,'" he said.

When his computer grades surpassed his marks in history, Kent switched minors. When they topped those he got in math, he made computer science his major.

As a master's student, he started tinkering with Java Virtual Machine. He modified networked computers to execute the program on several machines at once. At the PhD level, he worked with a supervisor to blend the program with hardware/software co-design.