Scientific Computing Research Environments for the Mathematical Sciences

Colorado School Of Mines, Golden CO

Investigators

Abstract

NSF Proposal DMS-0215491 PIs: Fairweather, Bialecki, Heremann, Rockwood, and Wang. ABSTRACT The new research computing facility in the Department of Mathematical and Computer Sciences (MCS) at the Colorado School of Mines will enhance its research computing infrastructure by providing faculty and graduate students with a state-of-the-art high performance computing environment. It will comprise 14 dual processor workstations, each with 1 GB memory and 40 GB hard drive, and a dual processor server with 2GB memory and 108GB hard drive. The machines will be configured with the latest version of RedHat Linux. These machines will be linked with some currently existing machines at MCS to extend the local distributed network for high performance computing purposes. Code development will be completed in the Linux/Unix environment with programming primarily in C/C++, FORTRAN90/95, and Java. In addition, MCS has site licenses for Mathematica, Maple, Splus, and Matlab to provide a full suite of symbolic computing capabilities. The new research computing environment is required to solve the complex problems being considered by researchers in MCS, specifically in the areas of matrix decomposition algorithms, nodal collocation for partial differential equations, development of symbolic software for nonlinear partial differential equations and lattices, data interpolation by finite element methods, and generalized least squares interpolation for modeling. Nine of the machines will form the nucleus of a computer laboratory to be used by graduate students assisting faculty members in their research. Under the proposed system, the machines can be used for individual computing by researchers, and also comprise a distributed computing network. The purchase of dual processor computers rather than single processor computers will double the number of processors in the distributed computing network at a cost increase of less than 20%. The distributed network will be used for code development, compilation, debugging, and production runs.