Emory Report
October 1, 2007
Volume 60, Number 6

Emory Report homepage  

October 1, 2007

Shared computing cluster now available for University

Donna Price is coordinator for communications and marketing services for Academic and Administration Information Technology.

On Sept. 3, the Office of Information Technology and the Research and Health Sciences IT division officially launched the Emory Life Physical Sciences cluster known as ELLIPSE, a 256 node, 1024 CPU, high-performance Sun computing cluster. ELLIPSE is the latest arrival in the Emory High Performance Compute Cluster (EHPCC), a subscription-based, shared resource for the University community that is managed by the High Performance Computing (HPC) Group.

The addition of ELLIPSE to existing campus computational resources, such as those at the Biomolecular Computing Resource (BIMCORE) and Cherry Emerson Center for Scientific Computation, moves the University into the top tier of institutions for conducting computational investigations in neural simulation, genomic comparison, biological sequence analysis, statistical research, algorithm research and development, and numerical analysis.

Computing clusters provide a reasonably inexpensive method to aggregate computational power and dramatically cut the time needed to find answers in research that require analysis of vast amounts of data. Eight and a half hours of ELLIPSE operation is equivalent to an entire year of 24-hour days on a fast desktop, and four to five days is equivalent to two months on its 128 CPU predecessor, which is still in service.

Dieter Jaeger, associate professor in the Department of Biology and chair of the executive committee that oversees the cluster, is one of the faculty who has used the EHPCC extensively for computational neuroscience focused on two brain structures, the basal ganglia and the cerebellum in the mammalian system.

“The new cluster is a central piece of what we want to do in research,” Jaeger said. “Computational neuroscience is an essential tool in understanding brain function. We could not handle the amount of complexity, the interactions of different processes in the brain, without a platform to recreate those dynamics in computer programs.”

Through the understanding of these processes and structures, better treatments can ultimately be designed for diseases like Parkinson’s, schizophrenia and depression.

About a year and a half ago, at Winship Cancer Institute, researchers at the Medical Physics and Engineering Research Laboratories had some problems they wanted to solve with high performance computing.

“For many years a lot of work has been going on about how much radiation the breast receives during mammography,” said Ioannis Sechopoulos, assistant professor in the Department of Radiology. “We weren’t able to find out how much the rest of the body indirectly receives from the mammogram. There were lots of estimates and calculations out there, but only one very limited scientific study.”

Analyzing data for research like this can take two, three or four years. The researchers created a scientific software program themselves using Geant4, a scientific simulation package developed by an international consortium of scientists, but needed to find the right technology environment to perform the computations.
“The choices were very limited,” said Andrew Karellas, professor of radiology and director of medical physics at Winship. “We could go out and spend $100,000 to develop a homemade mini-cluster, go out to a national laboratory and use something remotely, use the EHPCC, or we could abandon doing this research completely because it would be too expensive.”

They opted for the EHPCC and in two months had their answers, which are being reported in a series of four papers by Karellas; Sechopoulos; and Sankararaman Suryanarayanan, Srinivasan Vedantham; and Carl D’Orsi in the School of Medicine’s Department of Radiology. Two of the papers, that analyze a new three-dimensional breast imaging method, were published in Medical Physics in January and February, and two on radiation to the body from breast imaging have been accepted for publication in Radiology. All have the potential for very high impact.

“One paper that will have an immediate impact in the clinic was the study to find how much radiation the body receives from a standard mammogram using standard scientific methodology to simulate the human body and scientifically measure radiation on different body organs,” Sechopoulos said.

“I think it’s a phenomenal success story because of the overall chemistry between technology [and research], but more importantly the people behind it,” Karellas said. “We are very grateful to the administration and [Vice President and CIO for IT] Rich Mendola who, to their credit, recognized the need and invested in it.”

“We haven’t seen the end of it,” continued Karellas. “I’m very confident that there will be big successes from other groups who discover the new cluster.”

For orientation to ELLIPSE and help matching Emory’s HPC resources to computational research projects, contact Keven Haynes, senior manager, High Performance Computing; Steve Pittard, BIMCORE; or Jamal G. Musaev, manager, Cherry L. Emerson Center for Scientific Computation.

Donna Price is coordinator for communications and marketing services for Academic and Administration Information Technology.