Professors develop computer technology

Soon, a new innovation developed at Yale will make its way onto silver screens and even computer screens across the country.

Professors in the math and computer science departments have developed a new memory technology system that Pixar Animation Studios utilized during the production of their upcoming film “Cars,” which is scheduled to debut on June 9. The team, led by computer science professor Steven Zucker and mathematics professor Steven Orszag, is also currently working on the creation of a new data organization structure that will, for example, make Internet searches on more successful.

The team’s research breakthrough improves high performance computing (HPC) by integrating the storage and computing components of a computer cluster, which substantially increases the speed of the supercomputer, Zucker said.

Orszag said HPC involves the application of the most powerful computer technology to solve scientific problems in fields ranging from biology and chemistry to physics and engineering. HPC, he said, has become one of the three main pillars of science — along with theory and laboratory experiment — and is now widespread throughout the sciences.

But over the past decade, Orszag said, the growth in raw computing power has far outpaced developments in data accessibility.

“Over the last 40 years, computers have increased in power by a factor of about 100 million, but access to data has increased in speed by a factor of about 100 times less than that,” Orszag said. “Even in the last 10 years there has been a lack of improvements in access to data in comparison to the increase in computing power.”

The most powerful computers available, supercomputers, have also undergone changes within the last few years. Orszag said that over the last five years, there has been a shift from the use of monolithic supercomputers that fill entire rooms to computer clusters — hundreds and thousands of individual computers networked together. As of November 2000, only 20 of the 500 supercomputers in existence were computer clusters, Orszag said. But now, roughly 360 of the 500 supercomputers are clusters rather than monolithic.

“There has been a revolution in the way that scientific computing has been done over the past few years,” Orszag said.

The new development at Yale, utilized by Pixar, involves the improvement of such computer clusters. Orszag said the disk storage in a computer cluster is typically on a different server than the one used to carry out computations. But now, Orszag and his fellow researchers have integrated these two elements.

“It’s like thinking about how brains are organized,” said Zucker. “Human brains store an immense amount of information, and this information is stored on the same cells that process the information. That’s what we’ve done here; we have integrated storage and the use of information — in other words computing.”

Zucker said placing data directly on the computing unit, uniting memory and processing, makes the entire computer system faster.

The applications of this development involve improvements in input/output bound computing, Orszag said.

“This will help solve problems in I/O bound computing, like data mining,” he said. “Data mining is any sort of problem where you have to extract data from large data sets, like in bioinformatics, or with large computations in other areas like fluid mechanics and materials design. But these problems also arise throughout commerce and industry, so, for example, it can affect multimedia access and even movie making. The applications involve using this newfound capability of rapidly accessing very large data sets to address scientific problems that involve very large amounts of data.”

Orszag said a team of Yale professors from the mathematics, applied mathematics and computer science departments are currently working with this new computer cluster design to develop new mathematical methods to analyze and explore large amounts of data. This has the potential to help improve Internet search engines, Zucker said.

Search engines such as Google provide a common portal to help sift through the multiplying amounts of data available on the Internet, Zucker said. But, he added, computer scientists are still faced with the complexity of how to store, organize, retrieve and manage such a massive amount of information. The team at Yale is working on the development of a mathematical system that will allow for better data organization.

John Starks ’07, a math major, said that although he is unfamiliar with the particular mathematics of Orszag’s and Zucker’s new project, he thinks any effort to advance current search engine capabilities is appreciated.

“Google itself was a big improvement when it came out, and it was also based on new mathematic developments,” Starks said. “And certainly any further improvements are welcome.”