Acquisition of a High-Performance Parallel Computer for Mathematical Sciences and Applications

University Of Colorado At Denver-Downtown Campus, Denver CO

Investigators

Abstract

Acquisition of a High-Performance Parallel Computer for Mathematical Sciences and Applications The Center for Computational Mathematics (CCM) at the University of Colorado at Denver (CU-Denver) engages in mathematical sciences research and applications covering diverse areas of scientific computation, including model development, discretization methods, numerical linear algebra, optimization algorithms and applied statistics. Because the frontiers in each of these computationally intensive research areas continue to advance, it is essential to maintain high performance computing facilities capable of meeting ever increasing demands. The CCM is the only specialized site on campus for large-scale parallel computations. In addition to supporting research activities, the CCM also fulfills a critical role in the teaching mission of the University. Numerous advanced courses in Computer Science and Applied Mathematics rely heavily on the CCM computers. And the CCM facilities are vital to the success of our Mathematics Clinic Program. Research and teaching activities in computationally intensive areas are becoming increasingly limited by the CCM's outdated computational servers. There is therefore a pressing need to upgrade the computing capabilities of the CCM. This proposal aims to demonstrate the need for a new parallel computer by describing the following research projects and applications that will result from its use by faculty and students: Numerical Solution of Extremely Large Eigenvalue Problems; Acoustic Scattering, Domain Decomposition and Substructuring, and Distributed Parallel Programming; Parallel Eulerian-Lagrangian and Mixed Methods for Deterministic and Stochastic Subsurface Flows; Groundwater Modeling for Environmental Restoration; Validating New Mathematical Models for Swelling Porous Media; Numerical Solutions of Nonsmooth Equations; Estimation of the Value of Options; Atmospheric Models; Statistical Methodology for Massive Data Sets; Development of Accurate Finite Element Methods for Fluids Problems. We have identified the 64 processor Beowulf Intel-based cluster system to meet our computing needs. Each its 32 nodes has two PIII 733MHz processors and 2 GB of RAM. This powerful cluster with 64 GB total memory will handle not only the cross-disciplinary research described in the proposal, but will also be available to other users of CCM, including all faculty, postdoctoral scholars, visitors and students from the Department of Mathematics as well as affiliated faculty and students in the Department of Computer Science and the School of Engineering at CU-Denver.