INDUSTRY

CHAMPAIGN, IL -- The National Center for Supercomputing Application's (NCSA) Linux cluster based on Intel's 64-bit Itanium architecture is now online and available for use by the national academic research community. The cluster, named Titan, is the world's largest Itanium cluster, consisting of 160 dual-processor IBM IntelliStation machines. Titan has a peak performance of 1 trillion calculations per second (1 teraflop). It joins Platinum, NCSA's Pentium III Linux cluster, as the second teraflop cluster at the center. Both Titan and Platinum will later be incorporated into the new TeraGrid computing system. NCSA is working with three academic and several corporate partners to create the TeraGrid, the world's largest, fastest, most comprehensive distributed computing infrastructure for open scientific research. Funded by the National Science Foundation, the TeraGrid will include 13.6 teraflops of Linux cluster computing power distributed at sites in Illinois and California. "Titan's availability means that the user community now has an opportunity to deploy applications on Linux clusters with an expanded 64-bit address space and tune codes for future TeraGrid deployment,” said Dan Reed, director of NCSA and the National Computational Science Alliance, and a principal investigator with the TeraGrid project. "This will be a great resource for scientists who need a high-performance system that can analyze their data efficiently and allow them to get more science done. We believe Itanium family Linux clusters will be a major scientific computing platform in the years to come." Titan was purchased in 2001 and was put through a series of tests last fall and winter to determine its performance capabilities and ready it for general scientific use. Some scientists had early access to Titan during its "friendly user" period over the last six months, and several scientific codes have ported successfully to the system. These include Piecewise Parabolic Method (PPM), a code used to study hydrodynamics problems in science and engineering; Cactus, a parallel toolkit used in astrophysics and several other disciplines; NAMD, a molecular dynamics code used to look at processes in living cells at the atomic scale, GenIDLEST, a computational fluid dynamics code; and Tree Particle Mesh (TPM), a tree code that breaks massive problems down into many smaller ones. "Application tests over the last few months show the Itanium system to be extremely scalable relative to other high-performance computing systems," said Rob Pennington, associate director of NCSA's computing and data management divisions. "For many of these codes, it is important to have a system that can scale up and utilize hundreds of processors. Titan has proved it can do that." Visit www.ncsa.uiuc.edu for more information.