Cluster Computing for Mathematical Sciences at Texas A&M University

点击翻译按钮获取中文摘要

This proposal would provide computer resources to support three research programs in the Mathematical Sciences at Texas A&M University. Bangerth will work on the computational solution of non-linear inverse problems in tomography arising from biomedical imaging. Of particular interest is the optimization of experimental set-up that may involve tuning hundreds to thousands of design parameters. Carroll will work on nutritional epidemiology and genome-wide association studies. While statistical analysis of empirical data is feasible on a simple multicore machine, current methodology requires the performance of statistical methods to be tested on very large ensembles of simulated data. Sottile and Teitler will take advantage of cluster computing to compute local invariants of algebraic varieties, undertake a census of Galois groups of hundreds of thousands of geometric problems, and continue large-scale testing of conjectures in real Schubert calculus. The primary computer resources requested are 16 additional nodes for the 126-node Texas A&M University Brazos Cluster, which is optimized to perform the large number of single node/processor computations this research requires. This Brazos cluster is a high-throughput cluster created in the Summer of 2008, pooling the resources of five stakeholder research groups in four different colleges (Agriculture and Life Sciences, Engineering, Geosciences, and Science) and the Academy for Advanced Telecommunications and Learning Technologies at Texas A&M University. The Principal Investigators will become stakeholders in the Brazos cluster, giving them priority access to the new nodes and enhancing their access to the rest of the cluster.