Ter-Sacle Computing

Program Links

Laser-Plasmas

Quantum Computing

Signal Processing

Tera-Scale Computer

Gallery of
Fluid Motion

Dedicated to expanding
and disseminating
scientific knowledge,
especially in the
physical sciences.

Supercomputing Facilities...

The Institute has been designing cluster-style machines since the late 1990Õs. WeÕve built two machines: the first a 4-node Linux/Intel machine and the second a 4-node G4 machine. We are acquiring a 6- and a 10-node machine based on our previous designs, and we have the next-generation design on the table now.

The next generation machine has a scaleable design based around the 2GHz G5 processor, similar to but smaller than the Virginia Tech machine (put a link here) that came online earlier this year. We will build this machine in 3 phases over 2-3 years. Phase 1 will consist of a 2-rack configuration with a little more than 1/3 TFLOPS and 160 GB of memory at 1 GHz on a gigabit network. Phase 2 will add a third rack, more storage and a 20 gigabit switched network based on the Virginia Tech design. The third phase will be a new n-rack configuration based on then-current networking and chip technology in the 3-5 year time frame.

We have several user communities for our next-generation machine. One-third of the clock cycles will be used for Institute research programs in laser-plasma interactions and kinetic theory, quantum computing, signal processing, and our IR&D activities. One-third of the machine time will go to external researchers at various universities and research centers. And the remaining machine time will be devoted to education in and teaching parallel code development and performance profiling. We have developed an initial course around these ideas and are working to get the class listed in several university catalogs.

This project is a primary focus of our fund-raising efforts for this year. Phase 1 funding targets include equipment money and 5 years of operating capital. Phase 2 money will be smaller and go for equipment and to establish an operating endowment. Phase 3 will be the largest and go for the large machine and a permenant operating endowment.

Our approach to this ambitious project is low risk because it leverages our experience managing reasonably large projects at other organizations (Lockheed Martin, Aerospace Corporation, and Petrophysical Data Technologies), uses conservative technology upgrades, fills a need recognized by industry, and involves partnerships with many local educational institutions.

Combining this machine with our unique remote access and data warehousing technology will create a powerful resource for high-performance scientific computing research and an educational tool for future engineers and scientists. Phase 1 should be operational in the second quarter of 2005, ready for the fall semester.