University of North Carolina at Chapel Hill

Story Details

Established in 1789, the University of North Carolina at Chapel Hill (UNC-CH) is a leading research university. Many of the university’s health sciences groups work on National Institutes of Health (NIH)–sponsored projects to advance medical research in areas such as protein interactions, human-lung functions, and disease diagnosis. The groups typically use high-performance applications for modeling and simulation, but they wanted dedicated and specialized computing resources to speed research. Because their older system lacked the capacity for real-time analysis, scientists had to collect and model data in separate stages. They wanted a high-performance computing (HPC) system that would enable them to simultaneously collect and measure data and, as a result, make faster progress.

In 2005, representatives from seven research groups submitted a proposal to the NIH and requested funding for a high-performance biomedical image analysis system. The NIH agreed to fund the project two years later, and UNC-CH turned to Sun Microsystems to help design the new system. The university had used Sun products for more than 20 years and the proposal principal investigator was interested in the company’s recent HPC innovations. Although UNC-CH had considered other vendors, it decided that Sun could provide a more customized solution and better support for image-intensive applications.

Sun and UNC-CH spent most of 2007 designing the grid-based HPC architecture and began testing configurations in early 2008. When tests were complete, the university installed a factory-integrated system through the Sun Customer Ready Program. Because the entire system was already built, configured, and tested in a Sun facility before shipment, UNC-CH was able to cut hardware start-up time from six months to four weeks. UNC-CH formally introduced the new system to the research community in February 2009.

The UNC-CH grid — called the Biomedical Analysis and Simulation Supercomputer (BASS) system — consists of 17 Sun Fire X4600 M2 Servers, each with sixteen 2.8GHz AMD Opteron processors. An additional Sun Fire X4600 M2 was installed for backup, and another was installed as a file server attached to a Sun StorageTek 6140 array and a Sun StorageTek SL500 modular library system. The grid also includes 45 Sun Ultra 40 M2 Workstations with NVIDIA GeForce 9800 GX2 graphics cards and externally attached NVIDIA QuadroPlex Model IV visual computing systems. All the nodes run Red Hat Enterprise Linux version 5.0, except for the backup server, which runs Windows Server 2003 Enterprise Edition. The grid is networked by a Qlogic InfiniBand interconnect for low-latency massively parallel processing.

The university uses Sun Grid Engine version 6.1 software to tailor grid resources for different research activities. For example, it set up separate queues for projects that require large-memory systems, massively parallel processing, or graphics capability. UNC-CH also installed the Sun StorageTek Common Array Manager to administer the storage environment.

UNC-CH reports that the BASS system is easy to maintain and operate, with minimal training required for users. It is also reliable, with almost no downtime between monthly maintenance periods. But above all, UNC-CH researchers say that they are more productive with an HPC solution from Sun. “Users tell us how excited they are to be able to do things a hundred times faster than before,” says Russell M. Taylor II, research professor with a joint appointment in the departments of Computer Science, Physics and Astronomy, and Applied and Materials Sciences at University of North Carolina. “They can get work done to beat a deadline or get a paper out on time — it’s very gratifying.”