Web Content Display Web Content Display

LittleFe - An inexpensive, portable HPC education appliance

Many institutions have little or no access to parallel computing platforms for in-class computational science or parallel programming and distributed computing education. Key concepts, motivated by science, are taught more effectively and memorably on an actual parallel platform. LittleFe is a complete 6 node Beowulf style portable HPC cluster. The entire package weighs less than 50 pounds, easily travels, and sets up in 5 minutes. Current generation LittleFe hardware includes multicore processors and GPGPU capability, enabling support for shared memory parallelism, distributed memory parallelism, GPGPU parallelism, and hybrid models. By leveraging the Bootable Cluster CD project and the Computational Science Education Reference Desk, LittleFe is a powerful, ready-to-run, computational science, parallel programming, and distributed computing educational platform for the price of a high-end laptop.

Our poster describes the hardware and software infrastructure for the new LittleFe generation 4 units being produced this summer with funding from the Intel Corporation and the SC11 Conference. It also describes the parallel programming and distributed computing curriculum materials available for LittleFe/BCCD, including the software infrastructure which enables users to develop and test software on LittleFe/BCCD and then easily deploy that software on a variety of TeraGrid resources. We demonstrate that capability We will also have a LittleFe set up for people to examine and to run programs on.

Very Large-scale Biological Data Analysis using TerGrid Resources

This poster demonstrates the importance of TeraGrid resource utilization to analyze data from computational biology and biomedical informatics applications. The growth of genomic data in computational biology area has overtaken and outpaced both performance improvements of storage technologies and processing power due to the revolutionary advancements of next generation sequencing machines. The genomic data is doubling every 9 months, resulting in the exponential growth in recent years. On the other hand in biomedical area, the small carbon-based compound space with molecular masses in the same range as those of living systems for drug discovery is estimated to be 1060. There are millions of compounds that are available in various chemical libraries in the world. Currently there are around 22 million vendor compounds in ZINC and 33 million academic compounds in PubChem alone and this chemical space is growing very rapidly. However, data analysis becomes increasingly difficult and can be prohibitive, as existing bioinformatics tools developed in the past decade focus mainly on desktops, workstations and small clusters that have limited capabilities. Improving the performance and scalability of such tools is critical to transforming ever-growing raw data into biological knowledge that contains invaluable information directly related to human health. This poster describes novel software applications that include optimization techniques to improve the scalability of the most widely used bioinformatics tools on advanced parallel architectures, pushing the envelope of biological data analysis. We show that our improvements allow a near-linear scaling to tens of thousands of processing cores along with full machine capability runs on current petaflop supercomputers such as Kraken. These new tools, still under development, increase data analysis by four to five orders of magnitude and are helping foster research collaborations between two universities from Tennessee and four universities from South Carolina.

Pegasus WMS: Enabling Large Science on National CyberInfrastructure

Pegasus WMS is a workflow management system that can manage large-scale scientific workflows across grid, local and cloud resources. This poster will introduce the capabilities of managing these workflows on diverse national cyberinfrastructures like TeraGrid, Open Science Grid, and FutureGrid in an efficient, reliable and automated fashion.

The different national cyberinfrastructures that have been developed over the past decade offer different styles of high-performance computing. Leadership class systems, such as many TeraGrid resources, are optimized for highly parallel, tightly coupled applications. They provide scalable, shared filesystems like Lustre across the nodes on a resource. On the other hand, collaborative systems like OSG cater to high throughput loosely coupled applications. Typically, the sites don't provide a shared filesystem and encourage the model where jobs get their own data with them. Finally, cloud-based resources such as FutureGrid, and Amazon can be optimized for user's needs.

Pegasus WMS provides a means for representing the workflow of an application in an abstract form, agnostic of the resources available to run it and the location of data and executables. It then compiles these workflows into executable workflows by querying catalogs and sending the computations to the different resources using the Condor DAGMan as a workflow executor.

Pegasus WMS optimizes the execution as well as data movement by leveraging existing grid and cloud technologies via a flexible pluggable interface. While running on TeraGrid, Pegasus relies on the shared file-system to place input data for the workflows, while on OSG and cloud environments it may send input data directly to the worker nodes using Condor File I/O or S3 block storage. Pegasus also provides advanced features such as reusing existing data, automatic cleanup of generated data, job clustering and recursive hierarchal workflows with deferred planning. It also captures all the provenance of the workflow from the planning stage to the execution of the generated data, helping scientists to accurately measure performance metrics of their workflow as well as data reproducibility issues. Pegasus also interfaces with resource provisioners such as GlideinWMS or Wrangler to provision resources in advance of the workflow execution.

Pegasus provides debugging tools and monitoring tools that allow users to easily track failures in their workflows, by analyzing the underlying system logs.

Pegasus WMS was initially developed as part of the GriPhyN project to support large-scale high-energy physics and astrophysics experiments. Direct funding from the National Science Foundation enabled support for a wide variety of applications from diverse domains including astronomy, earthquake simulation, genomics, chemistry, biology, ocean modeling and others.

In the past year, astronomers have used the TeraGrid and Pegasus to generate an Atlas of periodograms for data from the Kepler space mission. The atlas will be published in the NASA Star and Exoplanet Database (NStED).

STAMPEDE: A Framework for Monitoring and Troubleshooting of Large-Scale Applications on National Cyberinfrastructure

Scientific workflows are an enabler of complex scientific analyses. Large-scale scientific workflows are executed on equally complex parallel and distributed resources, where many things can fail. Application scientists need to track the status of their workflows in real time, detect execution anomalies automatically, and perform troubleshooting - without logging into remote nodes or searching through thousands of log files.

In STAMPEDE, we have developed an infrastructure that captures application-level logs and resource information, normalizes these to standard representations, and stores these logs in a centralized general-purpose schema. Higher-level tools mine the logs in real-time to determine current status, predict failures, and detect anomalous performance. The STAMPEDE architecture consists of the workflow execution log collection, archival, and analysis components. Although STAMPEDE has been instantiated using Pegasus WMS as the workflow engine, our framework is designed to be extensible in order to accommodate other configurations.

For data collection, we have developed a program called monitord that continuously parses log files in real-time. The main task of monitord is to map incoming workflow logs to our general data model by generating NetLogger log events. The monitord program can insert the data directly into a SQL database, or alternatively, place the NetLogger log events in a message bus that is used to decouple the consumers of the streaming workflow data from the many possible clients. For this function, we can use the standard Advanced Message Queuing Protocol (AMQP), which defines an efficient publish/subscribe interface that is independent of the data model. STAMPEDE provides a NetLogger component that subscribes to the message bus and asynchronously inserts the data into a database. We are also helping to develop a component called Periscope that uses perfSONAR to gather, aggregate and cache data about network and computing resources relevant to the workflow. Finally, we provide an interface for external analysis tools to obtain raw or summarized data from the database, or directly subscribe to a stream of log messages for real-time data analysis. We have evaluated STAMPEDE using scientific workflows of various sizes from different applications, such as Broadband, CyberShake, Epigenome, LIGO, Montage, and Periodograms.

Open Gateway Computing Environments: Tools for Science Gateway Development

Science Gateways and portals are Web-based user interface and accessibility tools that provide user-centric views of cyberinfrastructure: they convert computing resources into tools for Web-based science and education. Although numerous production gateways have been developed, problems remain. How can operational gateways sustain themselves as underlying resources and middleware change? How can a gateway leverage modern open source and commercial Web techniques like gadgets and social networking within the current grid infrastructrue? How can a gateway wrap complicated science applications as robust services and workflows that really work in day-to-day operation? Can startup gateways reuse proven software from mature gateways and avoid reinvention and middle tier customization? In this poster, the authors present their efforts to address these problems through the National Science Foundation funded Open Gateway Computing Environments (OGCE) collaboration, an integrated group of software developers and operational gateway providers.

The goal of our collaboration is to provide high-quality implementations of software tools for Grid and Cloud-based scientific application management, workflow composition and enactment, resource discovery and fault tolerance, and social network-capable gadget component management. Going beyond these software tools, we investigate software engineering processes that support the full life-cycle of gateway software, from requirements gathering to operational use. Feature requests, enhancements, and changes to the software are managed using the Apache meritocracy model: the team is investigating long-term sustainability through participation in the Apache Software Foundation. Software developed by the researchers complies with relevant standards: scientific job management is provided through Web services generated by an application factory service; workflows are executed using open standards for enactment engines, and user interface components are compatible with the Open Social specification.

The poster will illustrate the OGCE project software engineering strategy along with the gateway software development tools and deployment highlights. The poster will introduce the OGCE Community driven software engineering and sustainability strategy through its participation in Apache Airavata and Apache Rave projects. We present OGCE tools in action in the GridChem/ParamChem (workflows), SimpleGrid (Web gadgets), and UltraScan (job management) projects. The investigators have also explored and demonstrated interoperability of the OGCE gadget components in the HUBzero environment and other gadget containers.

Modeling and Investigation of Disease Outbreaks in Hospital Networks with an Agent Based Model

Methicillin-resistant Staphylococcus aureus (MRSA) is a multi-drug resistant bacteria that is responsible for several difficult-to-treat infections, and represents a continuing public health and health care system problem. Hospitals and other health institutions within a given region are interconnected via patient sharing: changes in patient characteristics (e.g., habits or sociodemographic characteristics) or chronic disease epidemiology (such as with MRSA) in one hospital may affect others. We developed an agent based computer model that incorporates real life patient flow data and hospital characteristics as well as various empirical transmissions models (currently for MRSA). The simulation code is parallelized using MPI in order to facilitate fast and accurate extraction of pertinent statistics. The model enables us to conduct quantitative and detailed studies, such as potential regional outbreak scenarios and preventative intervention measures, with the premise of providing valuable input into public health decision making.

GPGPU-Based Parallel Viewshed Analysis on CyberGIS Gateway

Viewshed analysis is a well-known spatial analysis in Geographic Information Systems (GIS) that is often applied to determine the region of a terrain that is visible from specified locations. Many domain applications such as site selection and landscape planning require accurate viewshed on high-resolution digital elevation datasets. However, viewshed analysis of such datasets is computationally intensive and, thus, represents a significant challenge for conventional sequential algorithms to handle. The General Purpose Graphic Processing Units (GPGPUs) support massive data parallelism and provide a promising environment to resolve the significant computational intensity of viewshed analysis. This research describes the design and implementation of a GPGPU-based parallel viewshed algorithm, as well as its integration into the CyberGIS Gateway, a high-performance, distributed, and collaborative GIS gateway for cyberinfrastructure-based multi-disciplinary geospatial research and education.

Calculating the viewshed of a particular location involves shooting a separate line-of-sight (LOS) ray from the source (i.e. the location) to each target cell on a terrain raster in order to determine whether a target cell is visible from the source location. Consequently, viewshed computation on high-resolution raster datasets is both computing and data intensive due to a massive number of target cells to be checked. The design of our parallel viewshed algorithm aims to achieve two primary objectives: 1) mitigating performance bottlenecks caused by data transfer among disk, CPU memory, GPGPU memory and processors; and 2) effectively exploiting data parallelism on GPGPUs for data-intensive viewshed analysis. The first objective is achieved by two spatial domain decomposition strategies designed upon characteristics of both GPGPU architecture and spatial domain. A tiling-based spatial domain decomposition strategy is designed for loading input raster into GPGPU global memory, where each tile is made up of decomposed square blocks from input raster. The application of this strategy enables us to address significant data input/output (I/O) requirements and process sizable raster datasets which would not have been otherwise feasible due to limited capacity of GPGPU global memory. Because GPGPU on-chip shared memory has much lower latency than global memory, another strategy has been developed to improve memory access performance by further decomposing tiles into bins and loading bins into on-chip shared memory on stream multiprocessors (SMs) during viewshed computation. Because each target cell could be processed independently, massive data parallelism provided by GPGPU can be efficiently exploited by employing one thread to process one target cell and executing massive threads on all GPGPU cores simultaneously.

The performance of the developed parallel viewshed algorithm was evaluated on the Lincoln GPU cluster at NCSA on TeraGrid, which is equipped with NVIDIA Tesla10 4GB GPU cards and Intel 64 (Harpertown) 2.33 GHz dual socket quad core with 2GB ram per core. Our algorithm is implemented using the Compute Unified Device Architecture (CUDA) SDK version 3.2. Small datasets (less than 2GB) were used to compare the performance of a single-core CPU sequential algorithm [1] and our GPGPU algorithm. More than 100X speedup was obtained for a 1GB dataset on randomly selected viewpoint locations. Larger datasets, up to 12.9GB, were used to evaluate the effectiveness of leveraging low-latency on-chip shared memory. By loading further decomposed bins into on-chip shared memory, about 10X speedup was obtained than using global memory only.

Our algorithm has been integrated into the CyberGIS Gateway as a high-performance viewshed analysis service for community access. By taking advantage of the modular GISolve 2.1 toolkit, the creation of an integrated and interactive viewshed analysis Web interface is streamlined by using spatial middleware for job, data, and visualization management. For example, by using CyberGIS Gateway data access services, various digital elevation model (DEM) raster data such as LiDAR-based DEMs from OpenTopography (http://www.opentopography.org/) and the US Geological Survey, can be seamlessly located and transferred to TeraGrid for viewshed analysis.

New Features of the PAPI Hardware Counter Library

The PAPI specification and library have evolved from a cross-platform interface for accessing processor hardware performance counters to a component-based library for simultaneously accessing hardware monitoring information from various components of a computer system, including processors, memory controllers, network switches and interface cards, I/O subsystem, temperature sensors and power meters, and GPU counters. A new feature called user-defined events adds a layer of abstraction above native and preset events that allows users to define new metrics consisting of a combination of previously defined events and machine constants and to share those metrics with other users. One current effort is the development of a PAPI interface for virtual machines, called PAPI-V, that will allow users to access processor and component hardware performance information from applications running within virtual machines. PAPI continues to be widely used by application developers and by higher level performance analysis tools such as TAU, PerfSuite, Scalasca, IPM, HPCtoolkit, Vampir, and CrayPat. This poster will illustrate the new features of PAPI with examples from parallel application performance analysis.

Exploring mutational robustness as an emergent property in proteins using a high-performance computing approach

Robustness is a fundamental property of biological systems and can be interpreted as the system's ability to maintain function in the face of mutational or environmental challenge [1, 6]. In this work we present a distributed software framework that simulates mutational robustness as an emergent property in biological systems using extant protein structures and a pairwise contact model to calculate the Gibb's Free Energy of folding (_G). _G is a thermodynamic potential that measures the process-initiating work obtainable from an isothermal, isobaric thermodynamic system. We use a three-dimensional protein conformation to calculate the contact approximation for the effective folding free energy of proteins with known structure (_G), according to the method of Bastolla, et al. [2].

We assume a finite number of generations of a fixed population to which we apply an external mutation factor. Each individual consists of a single strand of DNA characterized by its _G of folding and can be regarded as a simple organism. Scwrl4 [3] is used to predict side-chain protein conformations for each mutated protein. This predicted three-dimensional conformation is used to calculate _G of folding of the mutant. We use the assumption that similar _G implies similar protein function. This assumption allows us for negative selection of individuals by measure of an upper and a lower threshold value where the individual survives when its _G value stays within that threshold once it has undergone random mutation. This type of mutation is said to be neutral because the protein can still fold and perform its function, thus keeping the organism alive [4]. Proteins that fall outside the thresholds are discarded from the population and allow room for the others to reproduce at random until population is filled again.

Let p be the number of individuals in the population, n the number of nucleotide sites and s the number of surviving individuals. In our simulation, mutational robustness is defined as the average __G=_Gpermutation-_Ginitial, or the difference between _Gpermutation of all the permutations of a fixed nucleotide (A, T, G, C) as single-site mutations of a DNA string and _Ginitial, or the change in Gibb's Free Energy of the initial DNA string. The calculation of the energy of an n-size DNA string is O(n2) and we need p of them for the selection of surviving individuals. Each single-site will have a permutation that is equal to the original DNA string (__G=0), which accounts for three distinct permutations per single-site. Thus, the mutational robustness of a single DNA string is found by calculating the sum of all __G divided by 3n. Each generation makes p+3sn calls to the energy calculation, thus the running time is O((p + 3*s*n)*n2). For our experiments p is set to 50, s is set to 1000 and n is set to 70. For this values of p, s, and n, a sequential implementation takes approximately 408 hours. A time profile of such sequential implementation reveals that 68.5% of all execution time is spent on energy calculations. This observation makes us realize that in order to improve the overall running time, we should seriously reduce the time spent on energy calculations.

Our initial approach is to use parallel computation to improve the overall running time of our simulations. It consists of three main efforts: (1) a Master-Worker approach using MPI to partition the DNA strings multiset and distribute a fixed number of strings to each of the workers available until all work is done. Although the asymptotic complexity of the energy calculation does not improve, this allows us to concurrently calculate _G for w strings, where w is the the number of workers available. (2) a shared memory model using OpenMP to improve speedup in the energy calculation itself. (3) a distributed hash table that maps [5] a string to a unique identifier and its previously calculated energy. Our preliminary results show nearly linear speedup on the parallelized version.

High-end computing for fluidized bed flow simulation

In a gasifier, the carbonaceous material undergoes several processes at high temperatures, including pyrolysis, combustion and gasification reactions. Fluidized bed technology is used in gasifiers because the process provides good mixing and promotes uniform heat and mass transfer between the gases and solid particles (e.g., granular material). There is usually a notable difference in the fluidization behavior between various solid fuel particles due to the particle characteristics, gas-particle interactions and especially the reaction kinetics. Research has been conducted on some aspects of the gasification process, but the physics and chemistry of reacting particles in a fluidized bed are less well understood. Therefore, it is important and timely to increase the fundamental understanding of particle mixing in fluidized bed gasifiers to help maximize the benefit of this technology.

Current research project aims to increase the fundamental understanding of particle mixing behavior and reactions in fluidized beds for solid fuel conversion using computational fluid dynamics (CFD). The overall goal of this project is to develop tools and methods to predict the operation and performance of gasifiers which will lead to more efficient gasifier designs. These goals will be achieved using in-house CFD code GenIDLEST which has the capability of Discrete Element Method (DEM) for getting accurate predictions for gasifier flows.

Traditionally the CFD simulation of the fluidized beds is limited to two dimensional (2D) flow simulations. The current work deals with the three-dimensional (3D) flow simulations of an industrial gasifier with 5.3 million particles. With the large number of particles the CFD-DEM coupled computations become very expensive in terms of computation resources. To achieve the most optimal performance, a scaling study of the multiphase flow problem is being performed on the TeraGrid resources of Nautilus and Ember. Also, comparative studies between 2D and 3D flow simulations are underway to better the flow physics understanding.

Future work for this project will involve inclusion of chemical kinetics and heat transfer in the computations.

Enhancing the Experimental "MATLAB on the TeraGrid" Resource

The "MATLAB on the TeraGrid" experimental resource has proven to be an important and unique parallel resource on the TeraGrid for computational science and data analysis. It has attracted many users new to TeraGrid and encouraged them to scale up their research problems. The resource provides seamless parallel MATLAB computational services running on Windows HPC Server 2008 to remote desktop (www.cac.cornell.edu/matlab) and Science Gateway (https://hubzero.org/resources/495) users with complex analytic and fast simulation requirements.

In a research partnership with NVIDIA, Dell, and MathWorks, Cornell is testing the performance of general purpose graphics processing units (GPGPUs) with MATLAB applications. MATLAB GPU computing capabilities include data manipulation on NVIDIA GPUs, GPU-accelerated MATLAB operations, and the use of multiple GPUs on the desktop via the Parallel Computing Toolbox and a computer cluster via MATLAB Distributed Computing Server. Testing is occurring on Dell C6100 servers with the C410x PCIe expansion chassis which supports server connections to NVIDIA Tesla 2070 GPGPU processors. In this poster, we will share information on system configuration, GPGPU testing results, and tips for testing and adapting codes for use with GPGPUs.

Dimming Wide Area Lustre Allocations Issues in Albedo

Deploying a wide area Lustre file system presents many administrative challenges amongst collaborators. Complications include user identification mapping, multisite firewall ruleset syncing, accounting and grant management, coordinated upgrades, and maintenance downtime.

Using the existing TeraGrid infrastructure has allowed these processes to become fully automated. Custom codes written from the Pittsburgh Supercomputing Center and Indiana University take advantage of the Teragrid Central Database for grant and account synchronization. Additional codes use the Teragrid Information Service for firewall propagation and creating user identification mapping tables for Lustre.

The end result is a fully automated wide area file system called "Albedo." Albedo's storage is spread out between six TeraGrid sites. The file system provides nearly 800TB of storage for our users.

Improving Coastal Predictive Inundation Models

Led by the Southeastern Universities Research Association (SURA), the Super-Regional Modeling Testbed involves twenty universities, eleven federal centers and programs, and two private corporations. This project is intended to benefit federal operational and applications entities and the academic research and graduate education coastal ocean science community. These benefits will be realized by developing a blue-print for the transition of promising models and skill assessment tools from an R&D environment to operational centers. The research-to-operations scenarios are closely guided by federal partners and beneficiaries including: NOAA/NOS/CSDL, MDL & Co-OPS; NOAA/NWS/NCEP; NOAA/NWS/NHC; NAVY/NAVO; NAVY/NRL; US Coast Guard; USGS; and USACE.

Numerical models which provide forecasts of weather-related natural hazards for residents of the US Atlantic and Gulf of Mexico coasts are being developed by various academic research groups, often as part of the U.S. Integrated Ocean Observing System (IOOS®). Assessing the relative accuracy, precision, robustness, reliability and efficiency of the models in use, as compared to each other and to actual data, has traditionally been difficult due to a lack of comparable model configurations, grids, forcings and consistent metrics for skill assessment. This NOAA IOOS-funded testbed is developing processes for skill assessment of models, with standard metrics that will facilitate transitioning to operations, scenario planning, event reconstruction, observing system design, and further model development.

Access to TeraGrid open science high performance computing systems is enabling the comparison of 2D to 3D models and the execution of coupled atmospheric and oceanic storm surge models over large physical domains (e.g., a few thousand kilometers) using high resolution spatial grids. Coupling of the WWM and SELFE models will significantly improve large-scale storm surge and inundation prediction. Any advances to these models for use in predicting coastal inundation also will pay dividends in their use in assessing water quality parameters and aiding resource managers to make better informed decisions about imperiled coastal waters.

Building community of practice through collaborative research: Modeling of nano-scale carbon and metalized carbon materials

We describe our effort to build a community of practice for the "EPSCoR Desktop to TeraGrid Ecosystems" project through a few collaborative computational material science research projects. Developing new computational methods that combine different Molecular Dynamics approaches in modeling of nano-scale carbon materials is a main theme of this effort. Our simulations include (a) quantum-chemical studies of a lithium-graphite-hydrogen cluster, (b) formation of new carbon nano-structures from graphenes, nano-tubes and fullerenes. The new development focuses on a quantum trajectory description for nuclei and quantum dynamical approaches for electrons. These projects are of high scientific impact and were chosen to exchange and strengthen the expertise in the following areas: (1) molecular dynamics, (2) quantum dynamics, (3) ab-initio simulations, and (4) multi-scale modeling. We discuss methods and results of simulations.

From 7 to 9 p.m. on Tuesday, July 19, there will be a poster session/visualization showcase and reception in Grand Ballroom Room E&F.





  • A Computational Model of Dynamic Social Entropy.  Andrew Pfeifer.
  • Visualization of M8 Earthquake Simulation. Amit Chourasia, Kim Olsen, Yifeng Cui, Thomas Jordan, Jun Zhou, Patrick Small, Daniel Roten, Geoffrey Ely, Dhabaleswar Panda, John Levesque, Steven Day and Philip Maechling.
  • Exploring Empty Space. Frank Willmore, David Simmons and Jack Douglas.
  • Blood Flow: Multi-scale Modeling and Visualization. Leopold Grinberg, George Karniadakis, Dimitry Fedosov, Bruce Caswell, Joseph A. Insley and Michael E. Papka.
  • Modeling Early Galaxies Using Radiation Hydrodynamics. Robert Harkness, Daniel R. Reynolds, Michael L. Norman, Rick Wagner, Mark Hereld, Joseph A. Insley, Eric C. Olson, Michael E. Papka and Venkatram Vishwanath. 
  • A Universe of Questions. Albert William, Michael Boyles, Chauncey Frend and Chris Eller.
  • A Novel Technique for Piezoelectric Microstructure Design. David Braun, Edwin Garcia and Carol Song.
  • Large-Scale Distributed GPU-Based Visualization Framework. Greg Abram, Byungil Jeong, Gregory P. Johnson, Paul Navratil, Kelly Gaither and Karla Vega.



The academic status of each presenting co-author is listed in parentheses. Best Poster awards will be given in three categories: high school, undergraduate, and graduate.

  • Plant Gene Expression: Gene Functionality and Timely Expression. Edward Folk (UNDERGRADUATE) and Lauren Bentley.
  • Plasmonics. Ian Reynolds (UNDERGRADUATE), Changjun Min, and Yorgos Veronis.
  • Computational Modeling and Design of Protein and Polymeric Assemblies. Christopher M. MacDermaid (GRADUATE STUDENT), Christopher Von Bargen (GRADUATE), William F. Degrado, Michael J. Therien, and Jeffrey G. Saven.
  • Use of Supercomputers for Theoretical Prediction of Two-Photon Absorption Properties of Organic Conjugated Polymers for Technological Innovations. Iffat Nayyar(GRADUATE STUDENT), Ivan Mikhailov, and Artem Masunov.
  • A REU Learning Framework for Cluster Challenge Training. Srirangam Addepalli, Jonathan Anderson, Xavier Bass (UNDERGRADUATE), Chirone Gamble Jr (UNDERGRADUATE), Jeff Pummil, Leon Smith, Jonathan Turner, and Margaret Warren.
  • Practical Investigation for Visualization Tools. Harsh Jain (GRADUATE STUDENT) and Hongmei Chi.
  • Bioinformatics Internship at PSC. Annie Kayser (HIGH SCHOOL) and Danielle Auth (HIGH SCHOOL).
  • A Computational Study of the Functionalization of Single-Walled Carbon Nanotubes with Macromolecules. Yinka Ogunro (GRADUATE STUDENT) and Xiao-Qian Wang.
  • The Remote Acquisition and Distribution of Oceanographic Data. Skanda Koppula (HIGH SCHOOL).
  • Enabling Large-scale Metagenomic Protein Family Identification on the NSF TeraGrid. Timothy Chapman (UNDERGRADUATE) and Ananth Kalyanaraman.
  • Prediction of Enantiospecific Oxidations by CYP2C19 Using a Chirality Code and Artificial Neural Networks. Jessica Hartman (UNDERGRADUATE), Steven Cothren, Jerry Darsey, and Grover Miller.
  • Non-local Interactions in Turbulent Flows: Departures from Classical Scaling and Intermittency Using TeraGrid Resources. Agustin Maqui (GRADUATE STUDENT) and Diego Donzis.
  • Mojave: An IDE for Cactus. Hari Krishnan (GRADUATE STUDENT) and Steven Brandt.
  • Artificial Intelligence and Skatterball. Jarrod Cingel (HIGH SCHOOL).
  • Molecular Parameter Fitting with a Genetic Algorithm. Robin Betz (UNDERGRADUATE) and Ross Walker.
  • Space Vector Modulation Voltage Source Inverter Variable Frequency Drive. Colin Hoylman (UNDERGRADUATE), Jeffrey Heck (UNDERGRADUATE), and Charles Hedrick.
  • Computational Investigation of Wireless Sensor Network Simulation. Dion Paul (GRADUATE STUDENT) and Hongmei Chi.
  • Exploring the Use of CUDA GPU Programming for High Performance 3-D FFT Applications. Alexander Arrico (UNDERGRADUATE), Yang Wang, and Roberto Gomez.
  • Computer Analysis of Plant Immune Response. Whitney Sanders (UNDERGRADUATE).
  • Computational Investigation of Available Resource Options for High Performance Computing. Robert Dunn Jr.(GRADUATE STUDENT) and Dr. Hongmei Chi.
  • Workflow and Job Management on the Grid. Daniel Dougherty (GRADUATE) and Tyler Eldred.
  • Numerical Investigations of Convection. Brandon Cloutier (UNDERGRADUATE), Paul Rigge (UNDERGRADUATE), Jared Whitehead, Benson Muite, and Hans Johnston.
  • Graphical Explorer of MPI Programs and Educational Plug-ins for Eclipse. Brandon Gibson (UNDERGRADUATE) and Ganesh Gopalakrishnan.
  • The Prickly Pear Archive. Dennis Castleberry (GRADUATE STUDENT), Oleg Korobkin, Steven Brandt, Erik Schnetter, Frank Löffler, and Jian Tao.
  • Intrusion-Detection in Large-scale Cyberinfrastructure Environments. Oliver Dillingham, Anenechi Egbosimba, Travis Jones (UNDERGRADUATE), Roy Blount (UNDERGRADUATE), and Jessie Walker.
  • Smoothing Seismic Velocity Model Using GPUs. Julio Olaya (GRADUATE STUDENT), Rodrigo Romero, and Aaron Velasco.
  • Preparing the OLCAO Program for Petascale Materials Science through Parallelization. Rachel Cramm Horn (UNDERGRADUATE), James Currie (UNDERGRADUATE), and Paul Rulis.
  • Methods for Automatic Preparation of Computing Environments for Application Execution: Experiences with MPI. Karolina Sarnowska-Upton (GRADUATE STUDENT).
  • Performance Analysis of Hadoop Jobs Using WebMapReduce. John Schudy (UNDERGRADUATE) and Baochuan Lu.
  • Caching in the Global Federated File System. Muhammad Yanhaona (GRADUATE STUDENT) and Andrew Grimshaw.
  • Wave Velocity and Frequency Spectra in Low-Temperature, Compressed 1-Dimensional Lattices. Meron Dibia (UNDERGRADUATE).
  • Gateway Optimal Resource Selection and Integrated Information Services Framework. Xuan Wu (GRADUATE STUDENT), Deivasigamani Suresh Kumar, Raminder Singh, Suresh Marru, and Marlon Pierce.
  • Metadata Management in Scientific Computing. Eric Seidel (UNDERGRADUATE) and Gabrielle Allen.
  • Anderson Localization of Phonons in Thermoelectric Nanostructures: A Path to Efficient Thermoelectric Energy Generation. Anthony Frachioni (UNDERGRADUATE) and Bruce White.