However much it excited cyber-psychologists, the “flash mob” phenomenon didn’t endure – largely, one would suspect, because it was so pointless. There are only so many times you are going to drop everything to join 50 other people who have had the same text message so you can all stand in a public place and go “moo” for 10 minutes.
Yet the ability afforded by technology for connected strangers to co-ordinate physical meetings, instant or otherwise, is a powerful one. The meetup.com website has played a significant part in the US Democratic primary campaigns. On another level, nzdating.co.nz and oldfriends.co.nz both attract more than 130,000 visitors monthly.
And recently, a group of students at the University of San Francisco (USF) held a cyber-rally with a purpose: to make a temporary supercomputer. As part of FlashMob I, about 660 computers, most of them laptops, were brought to the university’s gym, their owners having responded to an appeal on local computer bulletin boards.
The aim was to swiftly connect them to make one big computer, and get on the official list of the world’s fastest computers.
It has become relatively common – in part because it’s so cost-effective – to build supercomputers this way. Late last year, Virginia Tech launched Big Mac, a cluster of 1100 dual-processor G5 Apple Power-Macs, just in time to claim third place on the top 500 at a fraction of the cost of some of the conventional supercomputers it beat. A year before that, faculty and students at Massey University assembled Helix, a cluster of 132 AMD processors that outperformed New Zealand’s hitherto fastest supercomputer (the one at NIWA) and got us on the top 500.
But those were specially configured, identical PCs set up for only one job. How would this ragtag bunch of different machines play together? Not too badly, as it turned out. The organisers didn’t make the top 500, and they didn’t get near their target of 500 gigaflops (a gigaflop is a billion floating-point operations per second), peaking at 180 gigaflops and averaging 77. But it was a pretty good first outing.
FlashMob I emerged from a USF graduate class called “Do-it-yourself Supercomputers” and is the master’s thesis project of John Witchell, who described the result as “an ad hoc supercomputer – a supercomputer that’s built for a day or for a short period of time”.
Witchell and his classmates wrote the FlashMob software to harness the computers in six weeks, drawing on existing open-source software in the process, and then releasing their own work for use by others. As word spread, major computing facilities got in touch to ask them to track specific data over the day – no one had done this kind of thing before, with such a heterogeneous cluster of computers, and everyone was interested.
Witchell was not simply aiming to show that the job could be done quickly – he told reporters his aim was to get supercomputing “out of the laboratories and into the hands of the people. It used to be that you needed a double PhD to do this stuff. With the FlashMob software we’ve released, we believe that high schoolers can do this.”
It should be noted that that the FlashMob I team had the considerable advantage of being able to borrow top-of-the-line network switches to plug all the computers into a 10-gigabyte network, but it does appear that this really is DIY supercomputing, and will likely become more so.
So, apart from the undeniable geek appeal, what would be the point of lashing together a supercomputer? Well, groups of people could choose to assemble around laudable jobs that require serious computational power – specific problems in climatic research or cancer, for example, or testing an environmental impact model. Think of it as geek volunteering.
Then there are small film-makers, for whom pretty good PC-based digital effects are already a reality. What say a budding Peter Jackson and his friends could get 20, 30 or 100 souls together for a weekend of serious 3D effects-rendering at a temporary Weta Digital in the local Scout hall?
Or perhaps a temporary cluster could share the processor-intensive job of compressing MPEG video for DVD? Witchell is already working on a software module to do just that.
In the end, there’s no need to anticipate every possible use for this idea. It’s enough that a lab-coat computing model will, eventually, be turned loose on the street. And the street has always been good at finding its own uses for technology.