Interactive Supercomputing Inc. (ISC) was awarded a U.S. government agency research grant aimed at accelerating engineering and scientific discovery by pairing powerful parallel computers with easy to use desktop software.
The grant from Oak Ridge National Labs will be used to develop toolkits that parallelize the algorithms and models resulting from popular desktop technical computing software such as MATLAB, Python, R and IDL, also called Very High Level Language (VHLL) applications. The goal of the grants is to enable scientists and engineers to not only tap the parallel processing capabilities of high performance computers to solve huge computational problems, but to do so while minimizing development time. ISC was selected for the project due to its leadership in automatic parallelization technology that bridges desktop environments to high performance computers.
The Oak Ridge National Labs project, called "Towards a scientific data analysis platform for high level language application codes on high performance computers," will explore where it is best to parallelize codes written in a VHLL. Potential applications for this research include climate modeling, nanotechnology, biology and modeling fusion reactions. Codes written by the Lab staff will be analyzed to identify their scalability in terms of problem size and processor count. The project will utilize existing mathematical libraries and solvers on the servers, accessing them from a VHLL on the desktop.
"Scaling models and applications developed on the desktop to realistic datasets requires high performance computing servers," said Nagiza Samatova, senior research scientist at Oak Ridge National Labs. "This project would allow the researchers to tap high performance computers but still work in their interactive, familiar desktop environment."
The project will utilize ISC's Star-P(TM) software as the testing platform. Star-P is an interactive parallel computing platform that enables scientists and engineers to code algorithms and models on their desktops using familiar VHLLs and run them instantly and interactively on parallel servers or clusters. It eliminates the need to re-program the applications in C, FORTRAN or MPI to run on parallel systems. The projects will enable codes that typically "run out of steam" on desktop VHLLs to scale to very large data sets that are executed on parallel servers with many processors and large distributed memory stores.
"Considering the growing importance of parallel computing architectures, as well as the scientific reach of the Oak Ridge National Labs, this research should have a big impact in many areas of science, such as clean energy, environmental protection and national security," said Eckart Jansen, vice president of advanced development at ISC. "We look forward to working with the Labs in helping thousands of scientists and engineers who wish to access the power and memory of high performance computers without leaving their familiar interactive desktop environment."