Emory scientists will soon enjoy more computational muscle thanks to the University's acquisition of a new high-performance computational cluster. The 1,024 CPU-core cluster will significantly enhance Emory's existing computational resources, accelerating the pace of scientific discovery in a variety of fields including chemistry, biology, neurology, genetics, library science, pharmacology and medicine.
Anticipated to be up and running by June 2007, the new cluster will likely place Emory on the list of the world's 500 most powerful supercomputing sites.
Because scientists are generating and analyzing increasingly large quantities of data, powerful computational tools are crucial for those researchers and institutions wanting to stay on the cutting edge of research.
"This high-performance computer cluster will enable researchers to accomplish a variety of experiments -- using computer simulations -- which would prove impractical, impossible or too costly to do using conventional laboratory methods," says Richard Mendola, PhD, Emory vice president for information technology.
Researchers previously have had the options of investing in small-scale computer hardware and creating and managing their own computational clusters, using one of Emory's smaller, general-purpose clusters, or writing grant proposals to obtain financial support and access to one of the national supercomputer centers. "We see our new cluster as targeting the gap between the existing options at Emory and the very high end national supercomputer centers," says Dr. Mendola. "We feel there is real value added to the university by having a computational resource, particularly for junior faculty who may not have a track record in obtaining extramural funding for computational support. These researchers will now be able to access a well-subsidized resource--right here at Emory--through their own start-up funds," he says.
For Andrew Karellas, PhD, director of medical physics at Emory's Winship Cancer Institute and one of the world's leading experts in the development of new digital imaging detectors, the new computational cluster will allow him unprecedented research opportunities into early detection of breast cancer through the use of new imaging techniques.
Using the new cluster, Dr. Karellas, a Georgia Cancer Coalition Distinguished Scholar, and radiology faculty Ioannis Sechopoulos, will explore the level of radiation that patients receive from two new types of breast imaging techniques, known as digital tomosynthesis and computed tomography. Although not yet commercially available, these techniques may someday provide clinicians with more detailed views of breast tissue than ever before. In addition, several aspects of these techniques including improvements in image quality and three-dimensional visualization also need to be investigated. Dr. Karellas and his collaborators plan to explore these issues using the new cluster.
Dr. Karellas says it is important for clinicians to have accurate knowledge of how much radiation the breasts and other areas of the body receive during imaging procedures because this information helps determine the appropriate imaging procedure for each patient. With proper management of radiation dosage through accurate dosimetric information and good communication between medical physicists and physicians, the radiation dose and risks to patients can be minimized, he says.
"With the former cluster, some of the simulations we had wished to undertake would have taken several months and even years of computation time, making this type of research impossible. Computation that would have taken us months will now take no more than a few days. The new cluster will vastly widen our horizon in terms of new knowledge in this field," says Dr. Karellas. Like Dr. Karellas, researcher Andrew Jenkins, PhD, says he will use the new computational cluster to improve patient safety and treatment. Dr. Jenkins, an assistant professor of anesthesiology in Emory University School of Medicine, explores precisely how general anesthetics affect the central nervous system. Through rigorous computer simulations, Dr. Jenkins hopes to create safer and more effective general anesthetics.
"We do not yet really understand how anesthetics work at the molecular level. But an important part in better understanding these substances involves simulating how the structure and function of neurotransmitter receptors are affected by specific anesthetics?alone and in combination," says Dr. Jenkins. "The only way we can efficiently simulate the anesthetics' effect on the central nervous system is by using brute force computational methods, which the new cluster will allow us to do," he says.
The university initially will be offering researchers complimentary computer time to help them familiarize themselves with the new cluster, says Dr. Mendola. "After that, we're pricing the subscriptions so that the cost to faculty members will be cheaper than buying their own hardware or leasing it."
"The day we flip the switch, faculty will begin using the new cluster, and they'll be solving real problems with this new capacity. For them, the set up of the cluster can't happen fast enough," says Dr. Mendola.
Emory's High Performance Computing Group will be overseeing the installation and maintenance of the new cluster. Its computers were purchased from Sun Microsystems and its networking switches from Foundry Networks. The cluster will run the Linux operating system with a scheduler that will allow researchers to simultaneously submit thousands of processes. Researchers will be able to use various software applications depending on the type of research they are conducting.