The Department of Energy's Argonne National Lab has built its largest supercomputer system ever - a 256-node cluster that is to be based on Intel hardware and the Linux operating system.

Dubbed Chiba City for the futuristic city described in William Gibson's novel Neuromancer, the cluster will be open for use by researchers at other national labs, universities and even U.S. industry. Each node in the cluster is to be a dual processor server from VA Linux Systems, giving the cluster a total of 512 central processing units. IBM Netfinity servers will provide cluster management, file storage and data visualization capabilities.

Consultants from both IBM and VA Linux Systems' new consulting unit helped the lab in constructing the supercomputer, spokesmen for the two companies said.

Remy Evard, advanced computing manager at Argonne's computer science unit, said the two-day cooperative "barn raising" event used to build the supercomputer involved 50 Argonne scientists. The availability of tools for working with Linux and other examples of open source code makes it possible for many different participants to contribute to such a large scale project, Evard added.

Larry Augustin, president of VA Linux Systems, said Chiba City represents "a milestone in large-scale Linux systems design" and illustrates a software product that VA Linux has been working on practically since the company was formed, the VA Cluster Manager.

Chiba City is being demonstrated at Supercomputing '99, a supercomputing conference in Portland, Ore. Clusters typically require an interconnect or high-speed network to tie their nodes together, and VA Linux certified new high performance drivers for Linux to run on Gigabit Ethernet cards installed in each node.