Fast Track: USU Computer Science Grad Student Named Lawrence Scholar
Thursday, Feb. 10, 2011
USU doctoral student Steena Monteiro was recently selected for Lawrence Livermore National Lab's prestigious Lawrence Scholars Program.
Monteiro is researching techniques to manage information systems on par with the world's fastest computers, such as LLNL’s Dawn Supercomputer pictured above. Photo by Jacqueline McBride, courtesy of Lawrence Livermore National Laboratory.
Utah State University doctoral student Steena Monteiro, recently selected for the Lawrence Livermore National Laboratory’s prestigious Lawrence Scholars Program, is researching techniques to analyze and manage information systems on par with the world’s fastest computers.
IBM’s Sequoia supercomputer, set to launch this year at the California-based lab, is expected to deliver a scorching 20 petaFLOPs.
What the heck does that mean?
“FLOPS stands for ‘floating point operations per second,’” says Monteiro, who conducts research with faculty mentor Renee Bryce in USU’s Department of Computer Science. “FLOPS is a unit for measuring a computer’s processing speed. One PetaFLOPS is equal to a quadrillion FLOPS.”
A quadrillion? Looks like this: 1,000,000,000,000,000.
China’s Tianhe-1 supercomputer, currently the world’s speediest computer, clips along at a brisk 2.507 petaFlOPS. The U.S. Department of Energy’s Cray XT5 Jaguar at Tennessee’s Oak Ridge National Lab clocks in at a not-too-shabby 1.75 petaFlOPS.
So, 20 petaFLOPS? That’s roughly 20,000,000,000,000,000 calculations per second. Smokin’!
As a Lawrence Scholar, Monteiro will receive a fellowship of $53,000 per year for four years. Her proposed project, “Modeling and Detecting Performance Faults in Event-and Data-Driven Applications,” which she’ll pursue with LLNL scientists Greg Bronevetsky and Marc Casas-Guix, will form the basis of her dissertation.
“Steena is a wonderful student,” says Bryce, assistant professor of computer science. ”We are grateful that Lawrence Livermore National Lab has acknowledged the potential of her research by selecting her for this prestigious fellowship.”
Supercomputing applications comprise multiple components and are difficult to manage, says Monteiro, who completed an internship with LLNL’s Cyber Defenders program under Bronevetsky’s guidance in the summer of 2010.
“Today’s approach of manually managing large systems is already falling behind the complexity of today’s systems and will fail for the even larger systems of the future,” she says. “Our research approach toward successfully managing the complexity of large systems is to represent their behaviors as statistical models and use these models to infer the effects of various phenomena on a specific system’s behavior.”
The Sequoia supercomputer will serve the DOE’s National Nuclear Security Administration’s Advanced Simulation and Computing Program with huge-scale simulations.
“With computing at this scale, it’s difficult to detect whether a slowdown in performance is due to a security breach or a technical problem,” Monteiro says.
She and her research team will investigate varied parameters to determine which are most important to tune computer performance. The team plans to develop an automated metric collection system to pinpoint potential problems before they wreak havoc.
“We’ve been working on a model that can predict processor throughput up to two minutes in advance,” Monteiro says. “That’s a long time in our field. It was heartening to me to achieve this two-minute snapshot into the future.”
In addition to military simulations, supercomputers are used for large-scale financial applications as well as diverse scientific applications, ranging from genomics to modeling of weather systems.
Monteiro’s research could extend to future generations of sizzling supercomputers, including systems that will leave 20 petaFLOPS computers in the dust.
Contact: Steena Monteiro, email@example.com
Contact: Renee Bryce, 435-797-0737, firstname.lastname@example.org
Writer: Mary-Ann Muffoletto, 435-797-3517, email@example.com