ASCI's Debutante Debut

yoshi writes "Apparently, Lawrence Livermore Lab had an open house yesterday for ASCI White, the world's most powerful computer, and CNN has a story on it, including a picture of one of the sys admins! One of the great things about the system is how much information is available. Check out the hardware and software environments."

Sweet ride...

[Bowie J. Poag]

The last place I worked, IBM Storage Systems Division here in Tucson, had a bunch of these class of machines.. The test cell where I worked (for most of the time) was a roughly square area lined with 13 RAID arrays usually loaded top to bottom with 15K 72GB drives down one side, 3 IBM SP/1s (The same boxes that handed Kasparov his ass a year or two ago..) about 10 RS/6000's of varying horsepower, and a couple desks. One of the racks had exclusively nothing but 16-port Brocades in them, and what seemed like a mile of fibrechannel cable spewing out of it. I remember one slow day in particular, a friend of mine at work sat there, looked around, and tried to figure out how much money the company had stuffed in our little 9 by 14-square test cell in order for us to conduct our testing on the arrays... At $86,000 a piece, the Brocade rack was the priciest piece of real-estate in the lab, weighing in at $1.12 million dollars, or about $300,000 per square foot of floor space. Just within eyesight, we were encased within close to $20 million dollars worth of hardware, not including cables and the small stuff.

The Whopper may be King in the Land of Burgers, but IBM is God when it comes big iron. ;)

Nukes are just a front

[ssyladin]

A lot of people compilain about wasting all the cycles this monster can do on researching nukes. First, this thing would take a month to calculate the first tenth of a second - having to compute every quark, lepton, and electon in a nuclear bomb. At that level we already know (almost) any physics there is, so it becomes a huge billiard-ball problem.

Second - these machines only work on simulated nuclear testing for a short while (a few years). Then they go up for other "Grand Challange" problems, like immense weather calculation machines (who wouldn't want to know the exact minute it starts raining in your neighborhood?), particle calculations for the solar system or the galaxy, etc. We're up in arms about ASCI White, but what about ASCI Red, Blue, Mountain?

Furthermore, the research involved to build something like this benifits us down the line. Super-advanced routers, ultra-fast fail-safe network storage, improved networked processor topologies, distributed algorithms.