What with all the shuttling among Rockville, Cambridge, Palo Alto, La Jolla, etc., most frequent flyers in the genomics world don’t get much chance to visit Dallas. Thus it was with some anticipation that I arrived at the Dallas Convention Center right after Election Day. For there — wedged on the November exhibition calendar among The Great American Trucking Show, the North Texas Gun Club, and the Log Home Living Show — was SC2000, the annual Supercomputing Conference (www.SC2000.org).
ACM and IEEE sponsor this annual geek attractor, which draws more than 5,000 of the faithful to hear tall tales of teraflops and gawk at exhibitors pumping data at gigabits per second and up. (“Why, yes, my friend, that is uncompressed full-motion high-definition digital video you’re watching. And it’s coming right off the network. Look, no server!”) Bandwidth, processing, and high-tech swagger, Texas-style.
Long a showcase for academic computer scientists, DOE labs, and an assortment of DOD systems integrators whose lists of reference sites are all classified, the show has recently taken on more of a commercial flavor. Besides the big-iron supercomputer makers, the 150-plus exhibitors at SC2000 included other large computer companies, Linux cluster vendors, and the new breed of distributed Internet computing startups.
And one application on almost everyone’s mind was genomics. Stop by a booth, say the magic word, and watch the sales reps snap to attention. Many vendors claimed to have a handful of genomics customers already — and those that don’t sure want some.
Not surprisingly, Compaq was touting its highly visible installations at Celera and Sanger Centre, and Fujitsu was more than happy to discuss its role in Japan’s genome projects. A number of the DOE labs were showing off their genomics and bioinformatics activities. Even Cray (newly freed from SGI ownership) and SRC (the follow-on company Seymour Cray founded just before he died) professed great interest in the market.
But as Adrienne Burke reports in this month’s cover story, it’s the Beowulf Linux cluster world that may hold the greatest promise for the genomics supercomputing mass market. Cluster vendors API Networks, Linux Network, Microway, Paralogic, and Racksaver, all at SC2000, claim genomics customers.
And then there were the startup vendors in the so-called distributed Internet computing space: Parabon, Entropia, Distributed Science, and United Devices. They provide tools that allow the unused CPU cycles of all computers on a network to be harnessed for solving compute-intensive problems. Their grand vision is to harness all of the computers on the public Internet. A more modest vision is to allow organizations to harness all of the computers on their internal networks. What is among the proposed killer apps for all this processing power? Why, genomics, of course.
So supercomputing ain’t what it used to be. The take-home message from SC2000 is democracy. It has taken a few decades for revolution in the elite stratosphere of the computer world, but it has arrived. And unlike the applications that drove the development of supercomputers ¯ technical disciplines such as modeling nuclear explosions or ocean currents — genomics has the potential to help push this technology into the mainstream.
Dennis P. Waters, Chairman