Mob supercomputing

Print Friendly, PDF & Email

This isn’t news. In fact, since it happened in 2004, its the opposite of news. But I was intrigued by the idea of a “flash mob” computing event that was organized way back then and discussed last week in this blog post

Similar to cloud computing, there is another concept called “Flash Mob Computing”. As the name says, it refers to people bringing their computers to form an ad-hoc computer cluster trying to achieve supercomputer speeds. The idea as explained above is simple. People bring their laptops/plain old desktops, connect them to a network, form an ad-hoc computer cluster and achieve massive computing speeds.

The first ever such event was called, “FlashMob 1” and was held at the University of San Francisco Gym and was organized by Slashdot. The peak rate achieved was a mind boggling 180 GigaFlops (for the noobs : Flop = Floating Point Operations per second). The best complete run could achieve a rate of only 77 GigaFlops. The cluster ran on Morphix Linux. That’s nothing compared to the IBM Roadrunner’s 1.456 PetaFlops, but it still is awesome. The 1.456 PetaFlops is infact 20,800 times the stable speed achieved at “FlashMob 1”.

It is cool, but not exactly mind-boggling. In the summer of 2004 a system didn’t even get on the Top500 until it got over 624 GFLOPS on the HPL. But the idea as a social experiment is nifty. Organizing such a thing at SC’09 would be pretty swell. If one of you dear readers has the drive and time to do the organizing, insideHPC would sponsor it and help get the word out.