IBM makes it easier to build clusters of Linux servers

By Jaikumar Vijayan, Computerworld

IBM is making it easier for users to build clusters of Linux servers for running large computationally intensive applications.

The company today announced a series of prepackaged Linux cluster offerings that it claimed will allow users to tie up to 64, two-processor Intel Corp. servers into one massive Linux cluster.

Cluster configurations are used to harness the power of multiple servers to run large applications. Clusters are also often used to boost application uptime. For instance, each server in a high-availability fail-over cluster is capable of taking on the load of a failed server.

IBM’s Solutions Series for Linux Clusters combines its Netfinity PC server line with Arcadia, Calif.-based Myricom Inc.’s Myrinent cluster interconnect technologies, Santa Clara, Calif.-based Extreme Network Inc.’s Ethernet switches and terminal servers from Equinox Systems Inc. in Sunrise, Fla.

IBM will also bundle in supporting software and utilities for installing and managing applications, said David Gelardi, a director at IBM’s Deep Computing group.

Users can buy the clusters in configurations of 8, 16, 32 and 64 nodes. Pricing for an 8-node cluster with 16-processor support starts at $115,000, and will support the Caldera, Red Hat, SuSE and Turbo Linux distributions, Gelardi added.

The first Linux clusters are being targeted mainly at engineering and scientific markets because that is where the applications are, Gelardi said.

“Linux clusters have been very popular with the technical community,” said Rich Partridge, an analyst at D. H. Brown Associates Inc., in Port Chester, N.Y.

Not only do technical applications often need the high-end scalability provided by clustering, but they can also be manipulated more easily to take advantage of clustering technologies, Partridge said.

IBM plans to make similar clusters available for commercial applications such as data warehousing and Web serving in the future, Gelardi said. Also on the cards is a high-availability cluster configuration, he added.

The University of New Mexico, for instance, has already installed a 256-node Netfinity-based Linux cluster with more than 1.5T bytes of storage.

The system is being used by scientists and researchers around the country to run a range of biology, chemistry, physics, cosmology and other computationally intensive applications, said Dr. Frank Gilfeather, executive director of the university’s high-performance computing center.

One of the biggest advantages of running Linux is the fact that “it truly is the first environment that is upwards and downwards scalable,” Gilfeather said. “The same environment that runs on the desktop runs on the super-cluster,” without any changes.

“No other computing environment offers this advantage,” he added.

IBM’s latest Linux initiative builds on a series of similar moves during the past few months.

Just last week, for instance, IBM announced plans to release a new version of its AIX Unix operating system featuring extensive Linux support (see story).

In July, the company announced new hardware, software and pricing options designed to make it cheaper for users to run Linux applications on mainframes.

“Whether there is a huge demand for Linux or not, all the vendors are trying to make sure they don’t get left behind,” Partridge said.

“The development costs are not very large, so they can say they support it without [expending] a great deal of development resources,” he added.

David A. Bader
David A. Bader
Distinguished Professor and Director of the Institute for Data Science

David A. Bader is a Distinguished Professor in the Department of Computer Science at New Jersey Institute of Technology.