Software Investigations and Upgrades
In the past year, the GEaRS group made available many new software packages and updated many existing software packages. In total, the GEaRS group provides and maintains over 200 software packages that help meet the computational needs of faculty members in every College of the University. The following spotlight several of this year's most exciting additions.
Multi-Disciplinary Optimization Tools
MATLAB and Distributed MATLAB
X-ray Crystallography Tools
VASP
Collaborative Tools for Computational Researchers
Multi-Disciplinary Optimization Tools
The GEaRS group made a significant investment in state-of-the-art optimization software and services in 2005-06. Optimization products from ILOG CPLEX, GAMS Development Corporation and TOMLAB were made available to Penn State researchers. The software offerings have met the computational needs of a wide range of Penn State researchers associated with the Smeal College of Business, College of Engineering, and the College of Earth and Mineral Sciences. Specific faculty members with whom GEaRS collaborated and supported include: Dr. Terry Harrison (Supply Chain and Information Systems); Dr. Costas D. Maranas (Chemical Engineering); Dr. Antonios Armaou (Chemical Engineering); Dr. David A. Nembhard (Industrial Engineering); and Dr. Klaus Keller (Geosciences).
MATLAB and Distributed MATLAB
MATLAB is one of the most popular applications in the research community. It provides a simple interface to complex algorithms, to help researchers solve many diverse types of problems more easily; it is used across many academic disciplines. This year, due to faculty requests, the GEaRS group added 3 new MATLAB Toolboxes to the 14 toolboxes already offered. These included the Neural Network Toolbox, the PDE (Partial Differential Equations) Toolbox, and the Distributed Computing Toolbox/Distributed Computing Engine (together referred to as Distributed MATLAB) package. The number of licenses for the Statistics Toolbox was also increased from 5 to 30 to meet demand for this very popular Toolbox.
This investigation posed a particular challenge to implement seamlessly into the GEaRS group's existing computational infrastructure in an easy-to-use manner, but a novel solution was found, which worked so well, that it since has been shared with other universities for their own use. Having Distributed MATLAB available and easily accessible has allowed several faculty members to experience speed-ups in runtime that have reduced their time-to-results by as much as 99%. This is possible through the ability of Distributed MATLAB to take a problem that previously had to complete all of its computation on a single processor and split it to run across multiple processors on multiple systems.
X-ray Crystallography Tools
The GEaRS group entered a partnership with the Macromolecular X-ray Crystallography Facility (MXCF) in The Huck Institute of the Life Sciences to make X-ray crystallography programs available to the University community. This partnership leverages the respective domain expertise of each group by teaming the ability of the MXCF to identify and support the best X-ray crystallography programs with the ability of the GEaRS group to install and maintain the programs in the most optimized manner possible.
The first program to emerge from this partnership was CCP4, the program suite of the Collaborative Computational Project Number 4 in Protein Crystallography. This program was installed at the request of Dr. Neela Yennawar, acting director of the MXCF. GEaRS group members carefully modified the install process to allow CCP4 to use optimized math libraries available on the systems maintained by GEaRS. These optimized math libraries can give speed-ups of several hundred percent over libraries that are not optimized, and they are expected to reduce the time-to-results for researchers who use CCP4, as well as maximize the efficient use of the GEaRS group's computational resources.
VASP
The Vienna Ab-Initio Simulation Package (VASP) is a core package used by the Materials Science community at Penn State. In coordination with Dr. Jorge Sofo, associate professor of Physics and associate professor of Materials Science and Engineering, the GEaRS group made available optimized installations of the latest version of VASP, on all of its clusters. This involved not only compiling VASP for the different high-speed networks on the different clusters, but also running repeated benchmarks on each system using different compilers and different optimized math libraries to find the optimal combination that yielded correct results.
Collaborative Tools for Computational Researchers
One goal of the GEaRS group is to enable faculty partners to become more efficient at all levels of computation. Currently, the GEaRS group is working to create an environment that fosters a tight-knit community for researchers and their collaborators, both at Penn State and around the world. It is apparent that what defines collaboration is different for different research groups; therefore, most efforts will be involved in identifying common threads and tools that can help each group in their specific work and workflow models. Much of the GEaRS group's attention has been focused on providing cross-platform tools and pre-packaged toolsets, as well as frameworks upon which a mature toolset can be built.