MRI: Acquisition of Linux Cluster and Workstation Network for Biophysics, Mathematics, and Computer Science Research

Benedictine University, Lisle

Investigators

Abstract

A grant has been awarded to Dr. Ralph Meeker at Benedictine University to acquire a multi-processor Linux computer system and a network of computer workstations with stereo viewing capability for research in biophysics, mathematics, and computer science. Faculty and undergraduate students conducting biophysics research will perform molecular modeling simulations of protein binding and use the stereo workstations to analyze and visualize the results. Faculty and undergraduate students conducting mathematics research will use computer algebra software to study group cohomology and to make extensions to modular representation theory. The research in biophyiscs and algebra will provide data for Dr. Meeker and his students to measure the performance of the multi-processor computer system for both molecular modeling and computer-based algebra software as a function of the configuration of the computer system. Large-scale scientific computations, traditionally performed at reasonable speeds on specialized, expensive supercomputers or very slowly on engineering workstations, are being performed more and more on inexpensive Linux-based clusters of computers using a Linux variant to provide the cluster support. Rapid improvement in the performance of Pentium central processors, coupled with the stability and performance of the Linux operating system, has made it possible to assemble a networked cluster of personal computers at a fraction of the cost of a supercomputer. The equipment to be acquired with this grant includes a Linux multiprocessor computer system configured as an interconnected cluster with 16 separate computers. When assembling a Linux-based computer cluster for use in high-performance computing, one must take into account not only the raw computing power of the individual central processing units, but also system requirements in terms of memory and the network interconnection between the computers. This research project will measure the performance of the Linux-based cluster for both molecular modeling and computer-based algebra software as a function of the individual computers' memory configuration, the total number of clustered computers available for processing, the number of processors per computer in single-processor and dual-processor configurations, the number of simultaneous users, and what price/performance improvements are obtained from using dual-processor motherboards instead of single-processor motherboards. A network of computer workstations will provide researchers access to the cluster computer system. Three of the workstations will incorporate stereo viewing screens for visualization of three-dimensional molecular models of biological interest. Here is a brief description of the typical steps involved in this kind of modeling project. Identify a pair of molecules that are known to interact as a key step in a physiological process. If the molecules are large, and typically they are, the three-dimensional structure must be known or have a reasonable homology model from which to generate spatial coordinates and sequence regions of the molecules that interact. Results of x-ray studies conducted by other researchers can be used to determine this information. The molecules are then modeled by bringing them together, calculating the minimum of their molecular energy, and running through molecular dynamic calculations to identify possible binding orientations and their associated energies. For the mathematics research, the first stage of this research program will be analysis of data collected on the cohomology rings of the groups of order 64. The data analysis will yield insight on which groups of order 128 should provide "interesting" cohomology rings to compute. Results of this research will contribute to three areas of science: biophysics, mathematics, and computer science. The biophysics research is particularly useful in determining the binding of a fibrinogen chain to the integrin receptor of platelet cells, an essential step in blood clotting. The mathematics computational work should help distinguish irreducible abstract algebraic modules from other more complex modules by compiling a library of these basic algebraic building blocks. The computer science research will characterize the performance of a low-cost, high-performance computer system in computationally intensive applications to biophysics and mathematics. The involvement of the undergraduate students in these projects over the years have resulted in a number of undergraduate research papers and several who are now pursuing advanced degrees in science.