Stanley Watowich of the University of Texas Medical Branch in Galveston develops algorithms for a virtual experiment to test how more than 3 million molecules interact with 12 proteins related to dengue fever.
Grid supervisor Viktors Berstis, in Beaverton, Oregon, and his IBM team around the country fine-tune Watowich's code to make sure it will run safely and reliably on tens of thousands of different machines. Once it's approved, the grid team puts it in the work queue.
When there's space on the grid, Berstis's group uploads the program and corresponding data to the main servers in Toronto. The grid automatically assigns pieces of the project to thousands of available machines, based on each one's computing power.
Whenever the program senses that a volunteer's computer is idle, even for a moment, it borrows 60% of the computing power and begins crunching numbers in the background; each time the software resumes, it picks up where it left off.
When any one computer's work is done, the grid transmits the results to the servers in Toronto. Once all of Watowich's calculations are complete, WCG sends the results to his servers in Galveston, where his team looks for the most promising drug candidates to test further.
A version of this article appeared in the May 2010 issue of Fast Company magazine.