University of Pittsburgh

Reducing the Complexity in Complex Genetics

By Barbara Gibson

Dr. Michael Barmada, Assistant Professor in the Department of Human Genetics at the University of Pittsburgh, thanks his lucky stars that the National Institutes of Health waited before they awarded the grant that would expand the department’s Linux cluster from 48 to 250 processors.

“By the time the NIH approved our grant, our equipment quote was eight months out of date and NIH wanted a new quote,” Dr. Barmada says. “Apple had just announced the Xserve G5, so we suddenly had a new option. I looked at the Apple Xserve and said, ‘Absolutely — that’s where we want to be.’”

“We’re just a bunch of scientists. Were not programmers. We write code when we need to solve our problems. Mac OS X on the Xserves allows us to run our existing and legacy statistical genetics code — originally written for Solaris and Linux systems — without having to modify it extensively.”

The Case for Switching

Dr. Barmada’s team evaluated solutions from Dell, SGI, IBM and Cray and decided on the Xserve G5 because of its ease of management, the ease of porting UNIX applications to the Mac and “the added benefits,” he says, “that we can reap with the Xserve’s vectorization and G5 processor’s vector and 64-bit capabilities.”

“All of us in statistical genetics are on Macs,” Dr. Barmada explains. “We didn’t have any problem passing data between the Macs and the cluster, but managing the Linux cluster was a complicated task. Every time we wanted to add a new capability to the Linux cluster, we had to be very careful that we didn’t upset anything else.”

Moving to the Xserve also gave Barmada the environment he needed to support statistical genetics applications written in standard C code as well as legacy Fortran and Pascal code.

“We’re just a bunch of scientists,” Dr. Barmada says. “Were not programmers. We write code when we need to solve our problems. Mac OS X on the Xserves allows us to run our existing and legacy statistical genetics code — originally written for Solaris and Linux systems — without having to modify it extensively.”

Dr. Barmada says he’s just beginning to take advantage of the Xserve’s 64-bit architecture.

“Most of the work that we do involves linear algebra applications, and all of our programs include large matrix manipulations,” he says. “Once we modify some of our code to take advantage of the Xserve’s 64-bit processors and the Velocity Engine, not only will each calculation be faster, but we’ll be able to tackle even more tasks at the same time.”

He continues: “Initially, I thought of moving to the Xserve because of the easier administration. But it also gives us much more power and potential than we had before.”

1 2 3 Next Page >