Student design team helps university save $2 million annually in utility costs

A student design team in the Dept. of Mechanical Science and Engineering (MechSE) at the University of Illinois at Urbana-Champaign recently helped design a system for the new supercomputer keep its cool.

The Blue Waters petascale supercomputer is expected to be the most powerful computer dedicated to scientific research when it comes online in 2011. It also will require lots of cooling power, as its many processors generate vast amounts of heat.

"We could have easily have spent $20,000 to hire a firm to validate the original design, but we thought it would be nice to get some of our own engineering students involved in helping us solve a real problem at their alma mater," said John Melchi, senior associate director, administration directorate at the National Center for Supercomputing Applications (NCSA). "It was a great chance for undergraduates to work on such a national scale, visible project."

Four seniors in the MechSE senior design program—Stanley Berent, Ross Ratajczak, Andrew Thacker, and Christopher Theis—answered the challenge to cool the supercomputer. Their goal: To design a waterside economizer system with a cooling tower that would minimize lifecycle costs. Their system design is expected to save $2 million a year in energy costs.

While most supercomputing systems are air-cooled, the heat from Blue Waters will be cooled by the university's central chilled water system. The students—working under the guidance of senior design project coordinator Emad Jassim, developed a computer model to simulate the operation of multiple economizer designs involving different tower and component sizes on an hourly basis for an entire year using local weather data. They predicted the total annual chilled water reduction and total annual cooling cost associated with each design.

The students' analysis provided valuable guidance in the design process. It indicated that the annual operating cost of the system initially proposed by a Chicago-based consulting firm could be reduced by an additional $1 million. In the end, the design the students proposed was similar to the one proposed by the consulting firm—giving everyone involved a high degree of confidence that they had arrived at the optimal system. Once implemented, the cooling system is expected to reduce the annual chilled water usage substantially, saving NCSA a predicted $2 million a year in operating costs. At an estimated initial cost of just over $3 million, it is expected to pay for itself in approximately one and a half years.

According to the system designers, the larger the economizer system, which mainly consists of cooling towers, heat exchangers, pumps and piping, the greater the energy savings, but also the greater the initial cost. The optimum design lies where the sum of the initial and operating cost over the predicted life of the data center is a minimum, or has the least lifecycle cost. The team determined that delivering chilled water directly to the processors will avoid the inefficiencies associated with the intermediate step of using chilled water to cool air, and is expected to reduce energy consumption by approximately 40%.

When it is finally in operation, Blue Waters will give scientists and engineers the capability to process 1 quadrillion calculations per second at a sustained rate.








Enter CSE's ARC Awards! Entries due Aug. 29.