Lab to sample Linux for weapons work

[Eugen Leitl]

http://news.com.com/2100-1001-959082.html?tag=fd_top

Lab to sample Linux for weapons work

By Stephen Shankland
Staff Writer, CNET News.com
September 23, 2002, 4:50 PM PT

Los Alamos National Laboratory is buying a $6 million, 2,048-processor 
Linux supercomputer to run its nuclear weapons simulation software, an 
effort that will test the limits of these less expensive megamachines.

The lab has been a pioneer in building inexpensive supercomputers made out 
of ordinary computing components and the Linux operating system. Thus far, 
however, LANL's nuclear weapons simulation software runs on more expensive 
systems from SGI and Hewlett-Packard such as HP's $215 million "Q" now 
under construction.

A $6 million price tag may sound like a bargain in comparison, but 
software must be reworked to run using less expensive clusters of Linux 
machines. Though the new system will run unclassified programs such as 
predicting the properties of new materials, those tests will serve as a 
proxy to predict how well nuclear weapons simulation software works, said 
lab spokesman Jim Danneskiold.

The lab's central mission is ensuring that U.S. nuclear weapons will work 
as planned, despite aging and the current ban on actual nuclear tests. 
LANL has software that simulates the physical effects such as the extreme 
pressure and intense X-rays that accompany nuclear explosions.

Intel-based supercomputers are becoming less exotic, having escaped 
academia and found buyers in the private sector such as Companie General 
de Geophysique for oil and gas exploration work and MTU Aero Engine for 
engine design.

The new system, called the "Science Appliance" and built by Salt Lake 
City-based Linux NetworX, uses a cluster of 1,024 interconnected servers, 
each with two 2.4GHz Intel Xeon processors. It's a close relative to 
another cluster at LANL's sister laboratory, Lawrence Livermore National 
Laboratory.

The Science Appliance, due by the end of the year, will be capable of a 
peak computational speed of 10 trillion calculations per second, Linux 
NetworX said. The computing nodes will be stacked 50 to a rack, with 27 
racks taking up a patch of floor space about 18 by 25 feet. The nodes are 
connected with a high-speed switch from Myricom.

There are future expansion options in the LANL deal, said Clark Roundy, 
vice president of marketing at Linux NetworX.

There's a major difference compared with Livermore system, though: The Los 
Alamos machine has no hard drives. Instead, each computer fires up using 
software pulled over the network with the assistance of software called 
LinuxBIOS developed by LANL programmer Ron Minnich and others. LinuxBIOS 
also dramatically speeds the startup process to about two seconds, said 
Jason Lowry, Linux NetworX's product manager for cluster management tools.

Shunning hard drives cuts cost and power consumption, but more 
importantly, it improves reliability, Roundy said.

"If you think about what things are going to fail in a system, it's the 
hard disk or fan or power supply or something with moving parts," Roundy 
said.

Linux NetworX could benefit greatly from convincing the Los Alamos and 
Livermore labs that Linux clusters are worthwhile. The labs are funded by 
the Energy Department's Advanced Simulation and Computing program, which 
has spent hundreds of millions of dollars to advance supercomputing using 
machines made of comparatively inexpensive components.

The DOE program has underwritten many of the world's fastest computers, 
according to university researchers who monitor raw calculation speed at 
the Top500 organization. The program has underwritten Nos. 2, 6, 7, 9, 11, 
and 15 on the most recent ranking.