Virtually In The Fast Lane

College of Staten Island’s Athena, Zeus and Neptune make supercomputing even faster for research.

The Next Time you’re waiting at a toll plaza—frustrated by a line of cars blocking the E-ZPass lanes— take heart: Researchers Michael Kress and Jonathan Peters are trying to ease the congestion with the help of a sophisticated computer array that will be available to faculty across the University for high-end research projects.

The research team based at the College of Staten Island began developing computer simulations three years ago to better understand so-called “queue blocking” of the EZPass lanes at the Outerbridge Crossing tolls on Staten Island — tracking drivers’ behavior when they find themselves blocked. But to better understand the phenomenon, Kress and Peters needed a higher level of computational power that can crunch massive amounts of data.

Enter Athena, a high-speed computer with 96 nodes (each node has four processors) accompanied by Zeus and Neptune. This cluster of computers based at the college not only has given Kress and Peters a
formidable new tool, but also has helped ratchet up the research of other faculty and students. High-performance computers working in tandem, also known as simulation centers, enable researchers to
complete projects much faster—often in a couple of days instead of a couple of weeks—as well as to design problems and create “virtual experiments” that previously were impossible.

The E-ZPass “analysis helps understand how many cash lanes and how many E-ZPass lanes you need, especially during high-volume times like Thanksgiving and Christmas,” Kress said. “We can show at what point the system breaks down.”

Across the university, researchers are using high-performance computing to perform simulations in subjects ranging from molecular chemistry to studies of largescale weather and climate changes. This approach employs some of the complex equations used in traditional research methods, according to Kress, who also chairs the advisory board for CUNY’s Scientific Computation and Visualization Center. “But now we have so much computational power, we can mimic life.”

The advent of Athena is key to CUNY’s “Decade of Science,” the University’s renewed commitment to strengthening science, math, technology and engineering. “It’s essential to have a state-of-the-art
computational facility to take CUNY to the next level in terms of research,” said University Dean for Research Gillian Small.

University officials located Athena at Staten Island, Small said, since CSI had recently made high-performance computing a strategic priority for the campus—and had the necessary space, adequate electrical power and air-conditioning capabilities. “We still have clusters, but it makes sense to have one major facility,” Small said. Less expansive computational clusters already exist at the Graduate Center and at City College. Eventually, the goal is to link the University’s computer network to enable campuses in all the boroughs to take advantage of parallel computing facilities via desktop computers.

In the meantime, dozens of faculty and doctoral students use computational clusters at Staten Island, the Graduate Center and City College, in departments including chemistry, structural biology, economics, physics, engineering and applied math. What links many of these projects is the use of computation-based methods to imitate experimental conditions without having to recreate them in a lab.

At Staten Island, for example, Assistant Professor of Mathematics Andrew Poje is developing complex ocean models to determine where masses of particles—for example, oil spills—go when you drop them in the ocean.

“Ocean models are 15 to 20 years behind weather prediction,” Poje said. Conditions that may influence particle dispersion, such as rain, salinity, temperature and current velocity, generate “so much model data, it can take many days to figure out where many particles are going.” Supercomputing speeds the research. Instead of using one processor at a time to trace the dispersion of 100,000 particles over several days, Poje can use dozens of Athena’s 384 processors simultaneously, with each performing one of his computations. “The whole thing could be done in a day,” he said.

Poje acknowledges that the world’s oceans offer a vast arena for his simulations, but he hopes that in the next five years he can complete a more modest goal: creation of a good model of ocean dispersion for local coastal areas, such as New York Bay or Raritan Bay.

At John Jay College of Criminal Justice, Associate Professor Robert Till has been working on computer simulations of fire dynamics, examining combustion patterns and smoke production as fires break out in buildings or large public spaces. “Smoke is usually what kills people, so you want to run a lot of models to know where the smoke is going,” Till said. Such modeling is being used in the design and ventilation systems of new buildings and subway stations in New York City, he says.

Because Till’s work requires tracking dozens of variables and scenarios for each case study, CUNY’s high-performance computing facilities have made a critical difference. “Being able to run the model…on a certain processor in a couple of days—that’s an incredible resource,” he said. “It’s the difference between being able to do a project or not.”

Other researchers, like Anatoly Kuklov, use supercomputing
to explore the microscopic world. Kuklov, a theoretical physicist at CSI, focuses on quantum mechanics, particularly “superfluidity” and “super solid” states. With HPC simulations, he has shown that if Helium-4 is cooled to extremely low temperatures, these isotopes can demonstrate properties of a solid, but appear to act simultaneously like liquid under certain conditions— for example, if a defect is introduced into its crystalline structure.

These “virtual experiments” require large-scale calculations that researchers cannot perform on a simple desktop computer, Kuklov said. “It’s 100 times faster [with a high-performance computer]. Calculations that used to take a couple of years now take a couple of weeks.”

Besides speed and processing power, CUNY’s simulation center offers another important service: strong tech support. Much of that comes through Florian Lengyel, assistant director for research computing at the
Graduate Center, who has three technology fellows working with him to keep the network running and help faculty configure their software. “Florian is a huge asset,” Till said.

CUNY aims to expand utilization of the center throughout the university, while increasing computing capacity and storage. Faculty can visit for details on how to start up. Said Kress: “We have a sweeping vision: to develop a system to solve world-class problems in modeling and simulation and give the CUNY community at large this resource.”