IBM is helping the University of New Mexico build what it claims will be the world's fastest Linux supercomputer. It is a cluster of 256 rack-mounted, dual-processor IBM Netfinity servers running Linux, and it is able to execute 375 billion floating-point operations per second.
Later this year, IBM intends to market a packaged Linux cluster for commercial applications, officials said.
According to IBM, the machine, which will be called Los Lobos, will be the 24th-most-powerful computer in the world. It will cost just over $1.5 million. The clustering software is based in part on the open-source Beowulf code and other open-source projects, including IBM-developed software for installing, monitoring and managing clusters. Los Lobos is scheduled to be operational by this summer.
Dave Turek, vice president of Deep Computing at IBM, said Linux clusters are "still relatively immature in certain aspects" and that the company's RS/6000 SP massively parallel servers running AIX may be better adapted to certain demanding parallel processing tasks.
However, Turek said, Linux clusters will soon move well beyond the scientific market, where they have already established a strong foothold. He said that Linux clusters based on Netfinity servers are already being implemented at commercial companies by IBM's Global Services division. A packaged Linux cluster will ship from IBM well before the end of the year, he said.
Turek said that Linux clusters similar to the one being rolled out at the University of New Mexico can be used for transaction processing and electronic-business applications.
IBM already markets an 8-node Windows NT clustering technology called IBM Netfinity Availability Extensions for Microsoft Cluster Service.