Max Planck Institute at the edge of computational science: Prickly software tests new Intel processor
10. Oktober 2000
Cactus, a scientific programming framework and toolkit developed in Ed Seidel's research group at the Max Planck Institute for Gravitational Physics (Albert Einstein Institute, AEI), will be highlighted at the IntelTM eXCHANGE Expo in San Francisco, California, Oct. 11 and 12. Cactus will be displayed by the National Center for Supercomputing Applications (NCSA), running a numerical relativity simulation on a cluster of 16 Intel ItaniumTM processors.
Cactus was originally created as a modular, portable and high-performing framework for numerical relativists to develop application codes for studying such cataclysmic events as the collisions of black holes and neutron stars, or the collapse of pure gravitational waves. Now it's used by many research groups around the world. As Ed Seidel, head of the working group on "Numerical Relativity" at the AEI explains: "The latest release of Cactus provides a middleware layer for enabling not only numerical relativists, but scientists and engineers from any discipline, to easily develop parallel applications. Cactus is giving unprecendented performance on this next generation Itanium processor, and will enable many applications to scale from their desktop to the teraflop."
The name Cactus comes from the design of a central core (or "flesh") which connects to application modules (or "thorns") through an extensible interface. Thorns can implement custom developed scientific or engineering applications, such as computational fluid dynamics. Other thorns from a standard computational toolkit provide a range of computational capabilities, such as parallel I/O, data distribution, or checkpointing.
The Cactus Team at the AEI have been working with Intel and NCSA during the last year to port the framework to the Itanium architecture, and to optimize the code base to exploit the new processor to its full potential.
Intel's new Itanium processor has a 64-bit architecture which can provide the scalability and performance needed for high performance computing. In addition, the relative low-cost of these commodity processors and the availability of open source operating systems and software, makes clusters of Itanium chips an attractive option for research institutions and businesses alike.
The cluster on which Cactus will be demonstrated, consists of four machines, each containing four 667MHz Itanium processors. The processors are all running the open source Linux operating system, and use Myricom from Myrinet for interprocessor connections. Cactus will be demonstrated running in parallel on all sixteen processors, using the open source MPICH implementation of MPI for handling communications.
IntelTM is a registered trademark of Intel Cooperation.
ItaniumTM is a trademark of Intel Cooperation.