Browsing by Author "Teige, Scott"

Now showing 1 - 7 of 7

Acceptance Test Results for the Jetstream2 Production Environment
(2022-04-19) Hancock, David Y.; Turner, George; Weakley, Le Mai; Teige, Scott; Lowe, J. Michael; Bird, Stephen; Michael, Scott; Martin, Chris; Fischer, Jeremy; Snapp-Childs, Winona; Pierce, Marlon; Marru, Suresh; Wannipurage, Dimuthu
Building a Chemical-Protein Interactome on the Open Science Grid
(2015-03-15) Quick, Rob E.; Meroueh, Samy; Hayashi, Soichi; Mats, Rynge; Teige, Scott; Xu, David; Wang, Bo
The Structural Protein-Ligand Interactome (SPLINTER) project predicts the interaction of thousands of small molecules with thousands of proteins. These interactions are predicted using the three-dimensional structure of the bound complex between each pair of protein and compound that is predicted by molecular docking. These docking runs consist of millions of individual short jobs each lasting only minutes. However, computing resources to execute these jobs (which cumulatively take tens of millions of CPU hours) are not readily or easily available in a cost effective manner. By looking to National Cyberinfrastructure resources, and specifically the Open Science Grid (OSG), we have been able to harness CPU power for researchers at the Indiana University School of Medicine to provide a quick and efficient solution to their unmet computing needs. Using the job submission infrastructure provided by the OSG, the docking data and simulation executable was sent to more than 100 universities and research centers worldwide. These opportunistic resources provided millions of CPU hours in a matter of days, greatly reducing time docking simulation time for the research group. The overall impact of this approach allows researchers to identify small molecule candidates for individual proteins, or new protein targets for existing FDA-approved drugs and biologically active compounds.
Galaxy based BLAST submission to distributed national high throughput computing resources
(2013-03) Hayashi, Soichi; Gesing, Sandra; Quick, Rob; Teige, Scott; Ganote, Carrie; Wu, Le-shin; Prout, Elizabeth
To assist the bioinformatic community in leveraging the national cyberinfrastructure, the National Center for Genomic Analysis Support (NCGAS) along with Indiana University's High Throughput Computing (HTC) group have engineered a method to use the Galaxy to submit BLAST jobs to the Open Science Grid (OSG). OSG is a collaboration of resource providers that utilize opportunistic cycles at more than 100 universities and research centers in the US. BLAST jobs make a significant portion of the research conducted on NCGAS resources, moving jobs that are conducive to an HTC environment to the national cyberinfrastructure would alleviate load on resources at NCGAS and provide a cost effective solution for getting more cycles to reduce the unmet needs of bioinformatic researchers. To this point researchers have tackled this issue by purchasing additional resources or enlisting collaborators doing the same type of research, while HTC experts have focused on expanding the number of resources available to historically HTC friendly science workflows. In this paper, we bring together expertise from both areas to address how a bioinformatics researcher using their normal interface, Galaxy, can seamlessly access the OSG which routinely supplies researchers with millions of compute hours daily. Efficient use of these results will supply additional compute time to researcher and help provide a yet unmet need for BLAST computing cycles.
OASIS: a data and software distribution service for Open Science Grid
(J. Phys.: Conf. Ser. 513, 2014-06-11) Quick, Robert E.; Teige, Scott; Bockleman, Brian; Hover, John; Caballero, Jose
The Open Science Grid encourages the concept of software portability: a user's scientific application should be able to run at as many sites as possible. It is necessary to provide a mechanism for OSG Virtual Organizations to install software at sites. Since its initial release, the OSG Compute Element has provided an application software installation directory to Virtual Organizations, where they can create their own sub-directory, install software into that sub-directory, and have the directory shared on the worker nodes at that site. The current model has shortcomings with regard to permissions, policies, versioning, and the lack of a unified, collective procedure or toolset for deploying software across all sites. Therefore, a new mechanism for data and software distribution is desirable. The architecture for the OSG Application Software Installation Service (OASIS) is a server-client model: the software and data are installed only once in a single place, and are automatically distributed to all client sites simultaneously. Central file distribution offers other advantages, including server-side authentication and authorization, activity records, quota management, data validation and inspection, and well-defined versioning and deletion policies. The architecture, as well as a complete analysis of the current implementation, are described in this paper.
Performance Characteristics of Virtualized GPUs for Deep Learning
(2019-10) Michael, Scott; Teige, Scott; Li, Junjie; Lowe, John Michael; Turner, George; Henschel, Robert
As deep learning techniques and algorithms become more and more common in scientific workflows, HPC centers are grappling with how best to provide GPU resources and support deep learning workloads. One novel method of deployment is to virtualize GPU resources allowing for multiple VM instances to have logically distinct virtual GPUs (vGPUs) on a shared physical GPU. However, there are many operational and performance implications to consider before deploying a vGPU service in an HPC center. In this paper, we investigate the performance characteristics of vGPUs for both traditional HPC workloads and for deep learning training and inference workloads. Using NVIDIA’s vDWS virtualization software, we perform a series of HPC and deep learning benchmarks on both non-virtualized (bare metal) and vGPUs of various sizes and configurations. We report on several of the challenges we discovered in deploying and operating a variety of virtualized instance sizes and configurations. We find that the overhead of virtualization on HPC workloads is generally < 10%, and can vary considerably for deep learning, depending on the task.
SC|07 Bandwidth Challenge award-winning project: Using the Data Capacitor for Remote Data Collection, Analysis, and Visualization
(2007-11) Simms, Stephen C.; Davy, Matthew; Hammond, C. Bret; Link, Matthew R.; Teige, Scott; Baik, Mu-Hyun; Manri, Yogita; Lord, Richard; McMullen, D.F. (Rick); Huffman, John C.; Huffman, Kia; Juckeland, Guido; Kluge, Michael; Henschel, Robert; Brunst, Holger; Knuepfer, Andreas; Mueller, Matthias; Mukund, P.R.; Elble, Andrew; Pasupuleti, Ajay; Bohn, Richard; Das, Sripriya; Stefano, James; Pike, Gregory G.; Balog, Douglas A.; Stewart, Craig A.
In 2006, Indiana University led a team that received an honorable mention in the SC06 bandwidth challenge. The following year, IU expanded its team to include representatives of Technische Universität Dresden (TUD) in Germany and the Rochester Institute of Technology (RIT) in New York. The title of the 2007 project was “Using the Data Capacitor for Remote Data Collection, Analysis, and Visualization.” We believe that distributed workflows represent an important category of scientific application workflows that make possible new and more rapid discoveries using grids and distributed workflow tools. We believe that short-term storage systems have a particularly important role to play in distributed workflows. The IU Data Capacitor is a 535 TB distributed object store file system constructed for short- to mid-term storage of large research data sets.
Using the Data Capacitor for Remote Data Collection, Analysis, and Visualization
(2007-11-13) Simms, Stephen C.; Davy, Matthew; Hammond, Bret; Link, Matthew R.; Stewart, Craig A.; Teige, Scott; Baik, Mu-Hyun; Mantri, Yogita; Lord, Richard; McMullen, D.F. (Rick); Huffman, John C.; Huffman, Kia; Juckeland, Guido; Kluge, Michael; Henschel, Robert; Brunst, Holger; Knuepfer, Andreas; Mueller, Matthias; Mukund, P.R.; Elble, Andrew; Pasupuleti, Ajay; Bohn, Richard; Das, Sripriya; Stefano, James; Pike, Gregory G.; Balog, Douglas A.
Indiana University provides powerful compute, storage, and network resources to a diverse local and national research community. In the past year, through the use of Lustre across the wide area network, IU has been able to extend the reach of its advanced cyberinfrastructure across the nation and across the ocean to Technische Universitaet Dresden. For this year's bandwidth challenge, a handful of researchers from IU, Rochester Institute of Technology, and the Technische Universitaet Dresden will run a set of data-intensive applications crossing a range of disciplines from the digital humanities to computational chemistry. Using IU's 535 TB Data Capacitor and an additional component installed on the exhibit floor, we will mount Lustre across the wide area network to demonstrate data collection, analysis, and visualization across distance.