MRI: Acquisition of High Performance Computing Facilities for Research, Instruction and Outreach

Wilkes University, Wilkes Barre PA

Investigators

Abstract

This project aims to build and deploy a high-performance computing cluster to facilitate inter-disciplinary and collaborative research at Wilkes University. The cluster will support a broad range of computationally demanding research efforts including development of algorithms for computational chemistry and materials science, modeling protein folding and misfolding, predicting catastrophic geological events, large-scale analysis of bacterial genomes, and analysis of microwave tomography data for medical imaging, among others. Each of these projects will yield positive societal impacts including environmental hazard prevention, understanding and preservation of ecosystems, design of new nanomaterials, and detecting and treating diseases. These projects share the common theme of pattern recognition within large, complex, and noisy datasets and simulations. Thus, this instrument will enable a collaborative and multi-disciplinary approach for the execution of these projects while enhancing pedagogical activities within the college. The instrument will be administered so as to provide hands-on training for undergraduate students in modern computational techniques across multiple disciplines. The instrument will be the key piece of a new facility that will ultimately support the scholarship activities of a diverse group of investigators working mainly with first-generation college students as well as outreach activities aimed to inspire and encourage underrepresented groups and girls into scientific careers. The hybrid architecture of the instrument will simultaneously support data- and memory-intensive projects as well as real-time modeling and visualization. The primary computational engine, consisting of 8-nodes each possessing 24 computing cores (spanned across 2 processors), 4 commercial grade graphics cards, and 96 GB of RAM, will double as a visualization engine with the aid of virtualization software. This will allow the instrument to perform a variety of computational tasks such as molecular dynamics simulations, medium scale finite element analyses, training of smaller neural networks, performing medium-scale bioinformatics analyses, and MCMC analyses/ Bayesian modelling while enabling collaborative work among faculty and students as well as outreach activities through virtualization and visualization. Additionally, the cluster will include 4 ultra-high-performance nodes each containing 40 CPU cores, 1.5 TB of RAM, and 2 state-of-the-art GPUs. These nodes will facilitate extraordinarily demanding calculations such as the training of very large neural networks, simulating molecular systems with hundreds of thousands of atoms, and simultaneous alignment of multiple genomes. By using a combination of a standard batch queuing system as well as more modern container-based resource allocation schemes, this instrument will serve its ultimate goals of enabling computationally intensive, multi-disciplinary collaborative research at Wilkes University and providing hands-on experience and training for undergraduate students in key skills required for our nation's growing scientific work force. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.