Despite the threat it poses, IBM will soon begin offering support for Linux superclusters, according to an IBM official interviewed this week at the SP World conference. Such support may ultimately fuel faster adoption of Linux supercluster systems.
Once prevalent only in academic and government research, supercomputers have recently seen dramatic growth in the commercial sector, driven in large part by the explosive growth of the Internet.
In fact, more than half of IBM's sales of new supercomputers in the last two years have been to customers that use them as commercial Web-hosting servers, according to IBM.
While details of IBM's program weren't revealed, one executive said the decision is in keeping with the company's strategic initiative to support Linux.
Best of both worlds
"I see it as offering the best of both worlds," said Dave Gelardi, program director for business intelligent systems for IBM's RS/6000 brand.
While IBM may lose out on some opportunities to sell its multimillion-dollar system, the company can still profit from customers' decision to go with Linux.
Customers will still be able to develop their own relatively low-cost Linux-based system, he said, while assuring themselves of access to IBM's technical expertise.
While IBM may lose out on some opportunities to sell its multimillion-dollar system, the company can still profit from customers' decision to go with Linux.
"What I am providing here [as paid services] are skills, advice and counsel," Gelardi said.
While the company will initially offer support only for Linux superclusters using IBM Netfinity servers, Gelardi left open the possibility that IBM may expand the program.
The RS/6000 SP is basically a large-scale server designed to handle massive parallel workloads across hundreds or thousands of processors. The server operates using AIX, IBM's version of Unix. But the supercomputing power comes at a hefty price, with systems often costing millions of dollars.
Meet Los Lobos
In March, the University of New Mexico and IBM announced they had developed the first large-scale Linux supercluster system, nicknamed Los Lobos.
The US$1.5 million system, made up of 256 IBM Netfinity two-way servers running Red Hat Linux, could handle 375 billion operations per second -- making it the 24th-fastest supercomputer in the world, according to IBM officials.
While impressive, the development of Linux superclusters is still at an early stage, said UNM professor Brian Smith, co-director of the High Performance Computer Centre.
"This is not something that's ready for prime time," Smith said, "but it's not surprising to me that in a few months IBM will have units running able to do this stuff."
While admitting the Linux-based supercomputing system isn't as reliable as an IBM SP, Smith said it does have some potential advantages over the much more costly system. In particular, Smith said the Linux system offers the potential for easier expandability.
"We have used SP systems, so I'm familiar with them," Smith said. "Each time we added another node, we had to get another switch segment. Then very quickly we ended with a situation where the various switch segments weren't compatible. So you had to sort of get rid of all that equipment and start over again."
Has fewer limitations
But the Linux-based model avoids that limitation, he said.
"We can now grow in ways that don't require us to start over again," Smith said.
While companies could take advantage of information available on the Internet to build their own Linux superclusters, Gelardi doubts any businesses would take that risk.
"I've heard many CIOs and senior IT executives say, just on this point, 'How do I go to the CFO of this company and say, 'Well here's how we're going to get support. When we have problems we're going to throw them out on the Web and hope that somebody will be able to fix it.'
"I'm not sure you would want to be the one to bring that business proposal forward," he said.
