The companies are unveiling their new products Monday at the Supercomputing 2005 show in Seattle.
Linux Networx is introducing a new family of systems, the LS Series Supersystems, that includes a mid-range server and a high-end box that offers supercomputing performance.
The LS-1 is a modular system designed to combine performance with value, said Ben Passarelli, vice president of product marketing for the Salt Lake City company.
"Were emphasizing systems over cluster technology," Passarelli said. "We want to bring [to the systems] the price/performance of Linux clusters."
The system, powered by Opteron processors from Advanced Micro Devices Inc., can scale up to 128 nodes. The systems are put together on the factory floor—from bringing the hardware together to tuning the software stack—then delivered to the customers, enabling users to simply plug them in, he said.
The modular design also enables users to adjust their systems for custom applications. Linux Networx also offers a number of pre-tested subsystem modules that address such areas as I/O, visualization, storage and application acceleration that can be easily fitted into the LS-1 environment. The LS-1 will be generally available in the first quarter 2006.
The LS/X addresses the high-end of the market. The Opteron-powered system can scale up to 6,144 nodes and includes an integrated tier-1 switch architecture that reduces cabling while improving management, reliability and airflow, Passarelli said. It also leverages PathScale Inc.s InfiniPath I/O technology, a variant on the InfiniBand interconnect, he said.
The LS/X will be available to early access customers in December.
Both systems run Linux from Red Hat Inc. and Novell Inc.s Suse unit.
For its part, Penguin Computing is rolling out a new series of preconfigured Linux clusters. The San Francisco companys Application-Ready Cluster Portfolio come in three product families, all with bundled hardware and software and featuring its Scyld Beowulf Linux clustering product.
The clusters combine Penguins own Scyld Beowulf software and commodity technology such as AMDs Opteron chips and Intel Corp.s Xeon processors and InfiniBand interconnect technology.
The Portable Cluster, in a workstation form factor, scales from six to 24 CPUs that offer up to 200 gigaflops of performance and start at less than $20,000, officials said.
The Performance Cluster uses 1U (1.75-inches) and 2U (3.5-inches) rack-optimized servers to bring up to three-quarters of a teraflop of performance in a standard 42U (73.5-inches) rack. It can scale from 32 to thousands of nodes, and can use a variety of storage and interconnect technologies, including Ethernet and InfiniBand.
The High Density Cluster is a bladed environment that can run as many as 240 CPUs in a 42U rack.