For those of you who actually know something about science (as opposed to all you latte sipping yuppies who go on about “science” and how the nasty people on the right hate it, despite the fact that you graduated with a degree in English, or worse still, Sociology, from some pretentious liberal arts college), you know that by and large, scientific research in this country is funded by the US government. Now, certain applied fields in biology or chemistry are exceptions to the rule, but the majority of the time, most biologist and chemists are still lined up to suck at the teats of Uncle Sam, just like physicists, mathematicians and research-oriented engineers. Computer scientists are no different.
You might be asking yourself, how can this be? Isn’t all interesting work in CS done at Google Labs, IBM or Microsoft Research? The answer is: not on your life. As most scientists will tell you, corporate money is tough to get and is often very applied in nature. Companies (not unreasonably) are often reluctant to deal with university researchers since their goals are often contradictory (competitive advantage vs. publish or perish). So like everyone else, people in computer science line up for dosh from Uncle Sam. But while they can stand and wait on the National Science Foundation‘s doorstep with everyone else, computer scientists have a nice shortcut to dough through their sugar daddy at the Department of Energy.
“Big iron” computing (large scale, parallel) has, almost without fail over the last 40 years, been driven by the DOE. And that means it’s been driven by the DOE’s main goal (hint: it’s not alternative energy sources). Every year or so, Jack Dongarra and his friends at top500.org publish a list of the world’s Top 500 supercomputers, as determined by the LINPACK benchmark. As of the writing of this article (November 2006 List), 5 of the top 10 supercomputers are at DOE facilities and another is at France’s CEA, another weapons facility. The only two university computers in the Top 10 are in Spain and Japan. Machines at NASA and IBM (the only corporate machine in the Top 15) round out the Top 10.
By contrast, the fastest non-vendor (ie. not IBM or Cray) corporate machine belongs to COLSA, a technology services firm, sitting at #28. The next corporate machine sits at #46. There are seven more non-vendor (counting Intel as a non-vendor) corporate machines in the Top 100. This includes Merck and Eli Lilly, two mega-size drug firms. All told, non-vendor corporations run less than 10% of the Top 100 supercomputers. Why is that? First they don’t have the money to be on the cutting edge of computing. They’re more than happy to profit off of the technologies developed by IBM, Cray and the like for, uh, “other applications.” Corporate America cannot afford to finance the R&D of big iron computing and so it’s perfectly happy to let Uncle Sam foot the bill for the prototypes and buy their own scaled down versions once they’re cheap enough.
Now, none of this is necessarily a bad thing, but the next time you talk with a “big iron” computer scientist be sure (s)he realizes exactly who, or should I say, *what*, is paying the bills…