IBM and Canadian firm team up on Linux/Unix cluster

Continuing a string of announcements related to the development of high-performance Linux systems, IBM Corp. Tuesday said it’s building a server cluster that will combine the open-source operating system with Unix for use by a Canadian firm doing advanced medical research work.

The deal between IBM and Toronto-based MDS Proteomics Inc. is aimed at developing a supercomputer that can handle 700 billion floating-point operations per second and that will let scientists at the research firm conduct complex analyses of proteomics – the study of proteins and how they function. The two companies said they also plan to create a free online protein analysis database that will be accessible to scientists and other users from outside of MDS Proteomics.

Interestingly, the supercomputer will run Linux under the direction of a Unix server that will control the operations of the cluster, said Dave Gelardi, director of IBM’s Deep Computing division. The cluster will be built using 300 of IBM’s Linux-based xSeries 330 servers, assembled in three separate 100-unit nodes. Each server will contain dual 1-GhZ Pentium III processors and 2.2GB of memory, according to Gelardi.

The three cluster nodes will then be linked to an RS/6000 SP server running IBM’s AIX version of Unix. The Unix server will act as “an intelligent agent that dispatches work” to the Linux clusters, Gelardi said, adding that the machine will run an upcoming AIX release that’s supposed to have increased support for managing Linux applications.

One reason for the mix of Unix and Linux, Gelardi said, is that driver development and support for high-performance storage components is more developed under AIX at this point. “It’s a maturity issue,” he said. And those capabilities are important since the clustered system will include a massive data storage network based on IBM’s Shark disk array and DB2 database software, Gelardi added.

Among the applications envisioned for the supercomputer is research aimed at understanding how proteins interact and trigger chemical reactions in cells that cause diseases such as cancer, AIDS and depression. IBM and MDS Proteomics officials claimed that the machine would be the most powerful system in Canada and rank among the top 50 worldwide when it’s finished.

“Once you know the role that a protein plays in a disease, it is possible to develop drugs that target the protein and treat the disease,” said Frank Gleeson, president and CEO of MDS Proteomics, in a statement. The findings from research done on the new system could be used by pharmaceutical makers to develop new medicines, he added.

IBM will also make an undisclosed equity investment in MDS Proteomics, which is majority-owned by MDS Inc. in Toronto. The proteomics unit was founded two years ago.

The IBM/MDS cluster is just one of several high-performance Linux systems that have been announced in recent months for use in scientific applications. For example, the U.S. Department of Energy this month said its Sandia National Laboratories is teaming up with Compaq Computer Corp. and a biotechnology company to develop a US$150 million Linux-based supercomputer.

That followed several earlier IBM announcements, including a deal announced this month in which the computer maker and the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign said they’re building a pair of Linux clusters for use in advanced scientific and engineering applications.