Holy hardware!

High-performance computing cluster is UI's largest supercomputer ever

Tackling the world's biggest problems and advancing cutting-edge science and learning require computing power on a very large scale. Now researchers at the University of Iowa have access to a new high-performance computing (HPC) cluster that's the largest ever installed on campus.

"This system will enable University of Iowa researchers to make significant advancements in fields as diverse as biomedical engineering, flood prediction and mitigation, physics and astronomy, and global climate studies," says Mark Wilson, director of research computing at IIHR–Hydroscience & Engineering. "All these disciplines rely on the high-performance computing tools and methods provided by the new cluster."

Wilson cites two specific examples of cluster-enabled research:

• Eric Hoffman, professor in radiology, medicine, and biomedical engineering, and Ching-long Lin, professor in mechanical and industrial engineering and applied mathematical and computational sciences, have collaborated on the development of an innovative digital human lung model. The model combines the best in lung imaging capabilities with computational fluid dynamics to create a complete and complex digital lung to advance both research and treatment.
• Gregory Howes, assistant professor of physics and astronomy, studies turbulence in space and astrophysical plasmas and uses the HPC cluster for development and testing of complex, computationally challenging codes in preparation for runs on much larger National Supercomputer Centers elsewhere. He also is leading campus education efforts in the field of scientific parallel computing.

The HPC cluster—dubbed Helium—harnesses the computational power of 200 dual-socket, quad-core servers, yielding 1,600 computing cores. "Whereas ordinary servers often run at 5 percent of available capacity, computational clusters such as Helium tend to run at much higher utilization rates," says Ben Rogers, a systems administrator with the Institute for Clinical and Translational Science. "This higher level of utilization is due to a combination of the nature of scientific computing combined with specialized software that efficiently allocates scientific jobs to the available servers. What that means in layman's terms is that computational problems performed on a researcher's desktop machine that could take weeks or even months to complete can be run in days, hours and even minutes on Helium."

Helium's $1.16 million cost was paid for by a group of 12 researchers who realized they were better off pooling their research grant dollars and getting far greater computing power in return.

"The 12 individuals involved in the purchase of Helium had complete control over their money and could have individually built smaller clusters in their own labs," says Jerry Protheroe, interim director of research services for Information Technology Services (ITS) and ITS's coordinator of Helium's installation. "Instead, they were willing—enthusiastic, even—to share not just among themselves, but with others on campus who need a high-performance computing system to advance their research. There are other HPC clusters on campus, but none of this size. The next big breakthrough or cure for a disease could be here at Iowa—it could be because of this system and this collaboration effort."