We Came, We Saw, We Computed

I never thought I'd say this, but I participated in a flash mob last night (Sept. 15). We didn?t congregate on a street corner and start chanting or anything like that. Instead, we solved a molecular dynamics problem. ?Flash mob computing?, the brainchild of Patrick J. Miller, of Lawrence Livermore National Laboratories, is a way to build ad hoc supercomputers from unused desktop and laptop computers. Arrayed in the M. Carey Thomas Library at Bryn Mawr College were nine "slave" laptops under the control of one "master" working in tandem on a simulation of calmodulin unfolding upon interaction with anthrax edema factor. While Miller addressed the crowd of 50 or so chemists and students who came to see the event, our computers labored diligently. We didn?t break any new scientific ground, but we did solve 15,800 simulation steps (about 33 picoseconds of real time) in under an hour. We also, just for fun, calculated the first 15,000 digits of Pi -- that 15,000th digit is a 5, by the way. My Dell Latitude C400 never worked so hard. Key to flash mob computing is a custom-built Linux distribution (download) that runs off a CD instead of installing to the hard drive. All you have to do is boot from the CD, and viola, instant Linux. When you're done, remove the disc, reboot, and it's as if you were never there. Michelle Francl, a computational chemist at Bryn Mawr and Miller's sister, who participated in the event, has used flash mob computing in her own lab, hijacking four of her students' laptops to run molecular dynamics simulations too large for her own PC. "I can supersize my problems," she says. Life scientists can benefit from the approach too; Miller cites such applications as population dynamics and pattern matching. My wife, a biochemist, was smitten with the idea. She has trouble with multiple protein alignments and hopes her coworkers will donate their computers to the cause. She?ll need to optimize her algorithm for parallel computing to do that, however, and that takes savvy programming. Miller notes also that, because node-to-node communication is slow, problems in which one node?s work is dependent on another perform relatively poorly on such systems. If nothing else researchers can use flash mob computing to demonstrate to funding agencies that they actually can utilize supercomputing power, which will improve their chance of winning funding for time on massive systems like IBM's BlueGene/L, the reigning world champion. Says Miller: "It's a supercomputer on training wheels."

We Came, We Saw, We Computed

The Scientist ARCHIVES

Become a Member of

Meet the Author

Jeff Perkel

Related Research Resources

Bunsen Burners and Bad Hair Days

Research Resources

Podcasts

Webinars

Videos

Infographics

eBooks

Meet Honeybun and Breeze Through Viscometry in Formulation Development

Taking the Guesswork Out of Quality Control Standards

PFAS: The Forever Chemicals

Unlocking the Unattainable in Gene Construction

Products

Product News

Atelerix signs exclusive agreement with MineBio to establish distribution channel for non-cryogenic cell preservation solutions in China

Thermo Scientific™ Centrifuges with GreenCool Technology

Singleron Biotechnologies and Hamilton Bonaduz AG Announce the Launch of Tensor to Advance Single Cell Sequencing Automation

Zymo Research Launches Research Grant to Empower Mapping the RNome