Open source software proves affordable, flexible for NIH, DoD
by Colin Dodd
Ten years ago, engineers at the National Institutes of Health decided to bundle together a bunch of regular computers in order to create a supercomputing capability. The result proved successful and has grown since, enabling scientists to run large-scale computational experiments that would otherwise be impossible, such as processing thousands of DNA sequences or running six-month-long molecular simulations to study cell processes.
These simulations allow scientists to examine things that cannot be measured in a lab — for instance, how molecules permeate a cell membrane or how a drug interacts with a protein.
This system, known as Biowulf, has 6,500 processors communicating over a fast network and 8,800 gigabytes of memory. It would have cost the agency millions of dollars to buy enough software to make such a supercomputer possible. But the designers who built Biowulf in 1999 decided to use open source software, much of which can be obtained free.
“It has grown and grown since 1999, and we were able to do it in part because the software costs are almost zero,” said Steven Fellini, a systems specialist for NIH.
