Features:

MPI SOFTWARE TECHNOLOGY ENABLES NCSA'S NEW DELL CLUSTER

MPI Software Technology Inc (MSTI), a leading independent provider of high-performance middleware products and related services for parallel and cluster computers, announced that it is collaborating with Dell to build a cluster at the National Center for Supercomputing Applications (NCSA). Named "Tungsten," the cluster has a theoretical peak performance of 15.67 teraflops and ranks currently as the world's fourth most powerful computer as measured on the Top 500 List of supercomputers. MSTI, working as a Dell partner, is enabling the NCSA cluster through services that incorporate MPI Software Technology's Felix and ChaMPIon/Pro software. These packages contribute to the creation of a high-performance, production-computing cluster of dual-processor Dell PowerEdge servers. Tungsten will be used in research that includes studying the evolution, size and structure of the universe, investigating theories on the lifecycle of stars like the sun, modeling severe storms, studying the human genome and biological processes, advancing the drug design process and more. Furthermore, MSTI has obtained an HPL/Top500 benchmark result of 9.730 Teraflops on 2,464 processors of Tungsten with its ChaMPIon/Pro middleware, yielding 64.78 percent of peak efficiency. A further preliminary result of 10.12 Teraflops has also been achieved on 2,496 processors.

MSTI engineers have provided start to finish onsite support. By creating, benchmarking and validating the cluster the MSTI team has provided service and infrastructure support in helping finalize the hardware configuration. More importantly, their effort to deploy and commission the cluster will transition it rapidly to production computing soon after the conclusion of the Supercomputing 2003 conference in Phoenix.

MSTI's Felix aids in automating the installation, configuration, and management of the RedHat Linux Operating System, network configuration and software stack installation. By using Felix, MSTI engineers were able to perform these tasks in a remarkably short period of time. Felix software was able to perform the configurations on over 1,250 compute nodes at a time in a matter of a few hours. Felix enables "master nodes" of the cluster to support configuration and management of the "compute nodes" that are configured with Felix and the customer specific applications. The entire configuration is managed in a way to allow the customer access to parts of the infrastructure or all of it, while system administrators can easily update or repair any nodes that should need reconfiguration or restart over time. Because Felix management is based on a factory installed and supported Red Hat operating system build, the cluster nodes are preserved as fully functional independent computers. Using this technology, Dell and MSTI are able to deploy clusters of this scale quickly, reliably and with good ease-of-use factors after the deployment is complete.

MSTI's ChaMPIon/Pro is also being installed on the NCSA cluster. ChaMPIon/Pro is the first commercial MPI-2.1 version available for Linux. MSTI's line of MPI middleware products consistently outperform freeware by 10 percent to 50 percent, thus giving NCSA never before seen performance increases in a cluster this size. This contributes in important ways to the expected strong real-application performance of the Tungsten cluster, both for "legacy" MPI-1 applications, and for more demanding MPI-2 and multithreaded applications, including applications that perform massive I/O operations.

"NCSA is pleased to have ChaMPIon/Pro, a full MPI-2.1 implementation with commercial support, on its newest cluster. The software tools for MPI and cluster management offered by MPI Software Technology, Inc. will enhance the performance, usability and cost-of-ownership of this multi-teraflop cluster," stated Dr. Daniel Reed, director of NCSA.