Maximizing Research

By Keith Allison

Gunmetal gray pipes and poles crisscross the ceiling and erupt from the tile floor of the machine room. Toward the back of the room, bundles of colorful cables—technological sinews and arteries—cascade downward from a group of sleek black machines whose green LEDs flash and glow. Whirring fans cool and circulate the air, carrying the indescribable yet unmistakable scent of...computer processing. Close your eyes, and you can almost hear the hum of 512 2.2-GHz PowerPC 970 processors as they rip through a computation-heavy fluid dynamics model....

Meet Max, NYU's IBM eServer Blade Center supercomputer capable of a peak performance of 4.5 teraflops, making it the fastest computer in New York City and placing it among the 200 fastest supercomputers in the world.

The arrival of Max¹ and the innovative allocation of its resources position NYU to become a major center for supercomputing-based research across a broad spectrum of disciplines. Researchers involved with oceanographic and atmospheric research, and the study of biochemistry and the human genome, are already reaping the benefits of Max. They're discovering that not only does Max's computing power greatly enhance their ability to conduct research, but the policies regulating researchers' use of Max encourage an atmosphere of creativity and innovation that allows scientists to carry out experiments and run scenarios they might never have done otherwise.

Flood and Flow

NYU has long been committed to using technology in support of research.² As an example, in the 1950's, one of the first tasks for NYU's new UNIVAC was to execute and analyze model scenarios to predict what would happen to a specific region in the event of an accidental or intentional dam breach.

Fifty years later, the University's new supercomputer is swimming in similar waters, though with multiple magnitudes of additional processing power. Almost as soon as Max was installed in 2005, Assistant Professor Shafer Smith and his collaborators at the Courant Institute of Mathematical Sciences' Center for Atmosphere and Ocean Science (CAOS) began using the supercomputer to study ocean circulation and turbulence, running computer-intensive idealized simulations. The computer models Smith and his CAOS colleagues are running simulate changes in ocean flow and analyze how a tracer dye would be mixed in each region of flow. Their program uses 400 of Max's available 512 processors.

The oceanographic research being processed by Max could serve as a crucial key to further understanding how climate change—both man-made and natural—affects oceans, and how in turn the interaction of these two systems affects overall patterns of weather and life on Earth. "We're doing projects that, in some small way, help us better understand how the ocean works," says Smith, "and specifically may allow us to design better implementations of model physics in the major coupled climate model simulations that are run at national labs and used to assess future climate change."

CAOS Assistant Professor of Mathematics Olivier Pauluis uses Max to run high-resolution simulations of various atmospheric phenomena, such as clouds, hurricanes, and monsoons. He is also involved in researching how best to represent the behavior of clouds and convection in the next generation of global atmospheric modeling that computers like Max make possible. In addition to his own research, Pauluis teaches a graduate course on tropical meteorology in which the students use Max to analyze the behavior of atmospheric convection.

"The main advantage of computer models is that they let you study phenomena that you could not reproduce in a traditional laboratory environment, and at a fraction of the cost," says Pauluis, explaining why access to Max is important to his research. "For instance, while I could not possibly generate a hurricane in a lab, I can easily simulate one with a supercomputer like Max."

Professor Shafer Smith during a visit to the Max supercomputer facilities.

Although the Courant scientists were the first to tap into the computer's resources, a growing number of researchers outside of CAOS are harnessing Max's power. Researchers in NYU's departments of chemistry, biology, computer science, and astrophysics have recently begun using Max. Assistant Professors Fabio Piano and Kris Gunslaus of NYU's biology department, for example, are using Max to advance their genome research. The genome sequence for many species—including humans—has been successfully mapped, but the task now remains to understand how the products of the thousands of genes encoded in the genome work together to accomplish complex biological functions. Using functional genomic data from different experimental approaches, researchers are trying to understand how the cell works. To do this, scientists must devise a way to combine data so as to provide a robust picture. Max enables researchers to create dynamic models of the cell as it undergoes development and analyze these complex networks to identify global trends and explore local properties.

Innovative Control

For Smith and other scientists, Max's computing power is essential to their ability to run their simulations, but equally important is the University policy regulating access to NYU's new supercomputing center.

"There are other large parallel computers available at government labs," remarks Smith, "but these require strict limits on usage, and one must write very specific proposals to get access to them."

When NYU installed Max, however, the goal was to create a much more inviting environment for researchers. Max is a shared computing resource for NYU and NYU-affiliated researchers in need of massively parallel (MP) computing.³ Any NYU-affiliated researcher can get an account and run jobs, so long as processors are available. This friendly approach to administering access attracts scientists from both inside and outside of NYU.

Paul Maragakis, a professor in the Department of Chemistry and Chemical Biology at Harvard University, is using Max to conduct biochemistry research. He was attracted to Max and NYU in part because it's rare to find such a powerful computer with so few operational restrictions on the researcher.

"Our long term goal is to understand the dynamics and thermo-dynamics of conformational change in biomolecules. Conformational change, as embodied in biomolecular machines, is one of the fundamental manifestations of life at the molecular level," explains Maragakis. "The Max supercomputer has allowed us to test the conformational change formalism in a model biological system."

Smith and Maragakis agree that the availability of Max affords them the opportunity to take more risks, to explore avenues of research they otherwise might not have pursued, and to shorten the amount of time between the birth of an idea and its investigation. Thus, not only does Max's processing power fundamentally change the type of research that researchers are able to do, but the simple fact of Max's availability encourages a greater degree of experimentation and innovation, and allows them much more control over the computing environment. For Maragakis, this control comes in the form of flexible access policies that provide him with the unique opportunity to mix various types of parallel computations on one readily available platform, without the application and reporting delays sometimes associated with grants and access at other supercomputing centers.

Shafer Smith sums up what, aside from processing speed, might be Max's greatest contribution to research requiring supercomputing resources: "Researchers have no middle man to negotiate through in order to get access."

It is the vision of NYU that Max should continue to fulfill this role, to provide a truly researcher-centric supercomputing center and serve as the foundation of a larger supercomputing center. The work currently being done on Max demonstrates the computer's potential for providing powerful, high-speed resources in an accessible environment that encourages scientists to pursue innovative new ideas in innovative new ways.

Researchers who would like to request access to the Max supercomputer should send email to hpc@nyu.edu. Additionally, in her role as ITS' new Director of Academic Technologies, Heather Stewart is available to partner with NYU faculty as they seek funding for research with substantial technology components. She can be reached at heather@nyu.edu.

Footnotes

1. For more information about Max's arrival, see "NYU Acquires Fastest Supercomputer in New York City," by David Ackerman, in the Fall/Winter 2005 issue of Connect: www.nyu.edu/its/pubs/connect/fall05/ackerman_supercomputer.html.
2. Visit NYU's Information Technology Timeline at www.nyu.edu/about/techtimeline to browse a preliminary overview of some of NYU's IT accomplishments and to submit additional items.
3. Parallel computing is the simultaneous execution of the same task (split up and specially adapted) on multiple processors in order to obtain results more quickly. Systems with thousands of such processors are known as massively parallel. (http://wikipedia.org/wiki/Parallel_computing)

Author Biography

Keith Allison is a Technical Writer/Editor in the ITS Client Services Publications Group.