Grid computing efforts that capture media attention tend to be huge, powerful systems that measure performance in teraflops per second. But Grid computing's biggest benefits may be for very small operations with just 10-20 machines, according to Sun Microsystems' Peter Jeffcock.
"You can build a cluster Grid easily and cheaply, today, and get a great return on it," said Jeffcock, group marketing manager for Sun's Client and Technical Market Products Group. "You can grow toward something bigger over time, but you can do something on a very small basis today and get a big bang for your buck."
Duncan Ross and Darrin Perusich of Cognigen did just that. Perusich had tried other Grid products without success when he downloaded Sun's free Grid Engine software, "dropped it in the Grid, and got it working the way I wanted within 6-8 hours."
Cognigen, which conducts statistical analysis and modeling of clinical trials for pharmaceutical, biotech and healthcare firms, uses its Grid for its Fortran-based NONMEM (nonlinear mixed effects modeling) application. The firm's Grid, which is composed of Sun machines totaling 23 nodes, has "at least doubled compute power and cut costs by 90%," Perusich said. The system also lets the company queue jobs, making them easier to manage.
"This application helps us to get information out to clients faster," Ross said. And that helps life sciences companies reduce costs by making decisions sooner.
Sun's basic Grid Engine will remain free, but the company will charge for an Enterprise version it plans to release next year (it's currently available in a beta version). Jeffcock said the Enterprise edition adds a unique policy level for prioritizing projects. One benefit of that feature will be to encourage different cluster Grids to work together. "People might not want to share their CPU because they're afraid they won't get it back," he said. "This addresses that."
In some industries, Grids have become standard operating procedure, and companies that haven't adopted Grid computing are "behind the curve," Jeffcock said. Grids have become the norm in life sciences, scientific research and electronic design automation (EDA), he said.