Linux supercomputer on the way

The creation of the world's largest Linux supercomputer is clear evidence that the operating system (OS) is beginning to establish itself as a high-end server platform, according to IBM.

As part of a deal with Royal Dutch/Shell, IBM will develop a supercomputer using clustering to side-step some of the scalability shortcomings of the open-source OS.

The machine will run applications for number-crunching geophysical data collected in Shell's exploration surveys. It is expected to be running test software by the end of January.

The Shell deal is the latest in IBM's attempts to push Linux into the large-scale environment, and follows hot on the heels of a deal with Scandinavian telco and internet service provider Telia.

The main advantage for Shell is the cost, according to IBM. "In this case, a single $50 copy of Linux will be put across all 1024 processors, so the cost is truly tiny in comparison with proprietary Unix," said IBM European Linux solutions manager Nick Davis.

Wide-scale deployment of Linux on enterprise systems is being held up by scalability problems, according to analysts, who say Linux cannot scale beyond two processors without complex clustering technology.

"Larger Linux systems may prove better than either mainframes or Unix systems, but Linux still needs to be tested running real applications in the enterprise environment," said Chris Ingle, a consultant at researcher IDC.

But Davis said that the answer to the scalability problem is just around the corner. "The Linux 2.4 kernel is in final beta testing now and will overcome the issue of scalability without using clustering technology," he said.

"There will be little doubt about Linux's capability in the commercial environment by the end of 2001 and it will be the dominant server operating system within three years," he added.