High-performance supercomputers provide significantly increased memory and computing power, while allowing for expanded applications that can advance research and development in areas such as astronomy, medicine, and engineering. However, these systems use a tremendous amount of energy and produce a lot of heat, which requires the use of additional high-power cooling systems. As the benefits these supercomputers provide become more essential, methods are being developed to both reduce energy use and operate these systems more efficiently.
Currently a research team at RIT, led by research scientist Dr. Lizhe Wang, has several efforts underway designed to reduce energy use in supercomputers and create more efficient procedures for the operation of these systems.
"Supercomputers represent the next step in advancing computing application and development," notes Wang, assistant director of the Service Oriented Cyberinfrastructure Lab in the B. Thomas Golisano College of Computing and Information Sciences. "However, the energy use associated with these systems, and the impact this has on the environment and system cost, is currently an impediment to expanding supercomputer use."
"At RIT, we are attempting to develop processes that can assist in reducing run times, better schedule jobs, and reduce the need for cooling systems that add to energy use," adds Wang.
For example, the team is building a Green Web 2.0 portal for use by supercomputing centers. The portal will allow researchers from all over the world to access grid resources in the most energy-efficient mode possible. The portal will also allow computer centers to better schedule use of its supercomputers and more efficiently operate the entire system.
"One of the main problems with supercomputing centers is the large number of users and difficulty in efficiently and productively scheduling the literally thousands of jobs the system is required to run," Wang says. "By creating a virtual portal and manager, more people can use the system more efficiently."
The work conducted is partially funded through the National Science Foundation.