SGI, NASA reach server milestone
NASA has successfully fired up the first SGI Origin 3800 supercomputer with a single operating system spanning all 512 processors.
Though NASA has earlier Origin 2000 systems of this size running, this is the first time the new Origin 3000 series has made it to the operating system milestone. In addition, NASA hopes to get a single operating system running on a 1,024-processor system this June, spokesman Greg Slabodkin said.
It's difficult to get an operating system to span such a large system. But doing so makes it easier to write software than with some supercomputer designs, which essentially have several operating systems running in parallel and must coordinate operations such as communicating what data is stored in what patch of memory.
Though SGI's designs simplify these memory issues, it takes complicated and expensive customized chips, as well as a specially written operating system, to do so. SGI's technology is called non-uniform memory access (NUMA), which refers to the fact that it can take a different amount of time to write information to memory depending on how far away a memory bank is from a CPU.
NASA's Ames Research Center is using the system for calculating airflow around aircraft, simulating life's origins and predicting hurricanes, among other tasks, SGI said.
SGI's current system uses its own MIPS chips and its IRIX operating system, but the company is adding Intel's much-delayed 64-bit Itanium chips to its product line as soon as Intel releases the chip later this year, Jan Silverman, vice president of advanced systems marketing, said in a statement.
The Intel systems will run on Linux.
SGI has tried for years to move beyond its specialties--scientific and technical computing and digital content creation such as animation. But though analysts have viewed the Origin 3000 systems favorably, they have yet to achieve the success in more mainstream business accounts as competing Unix servers from Sun Microsystems, IBM, Hewlett-Packard and others.