SC10 Education Program Curriculum Fair Posters
Call for Abstracts
The SC10 Education Program is pleased to announce the opening of abstract submissions for the 2010 SC Education Program Resource Fair. This event allows the education community to share and find good work among the community of high-performance computing and computational science educators present at SC10.
We encourage you all to share your educational resources, programs, research and student projects with others having similar interests and facing similar challenges. While the most common presentation format will be in the form of posters, other approaches to sharing your work are also encouraged (though you will need to provide your own infrastructure for projection etc., wireless connectivity will be available).
Please use the following link to submit your abstract to participate in the program - submit. The main resource fair event will take place Saturday evening from 5-7 pm at the convention center. This will be followed by an SC Communities party until 10 pm. Posters will remain on display throughout SC10.
We have 40 4'x8' display boards available to us, though we would recommend that you make your posters 4'x4' to make transport easier. In addition, there will four tables with power drops (four per table) to allow for media presentation as well. Presenters will need to provide their own presentation equipment. Those who do bring equipment will be able to store their equipment in the education office after the resource presentation session at 7pm so that they can go directly to the communities event at the aquarium. Equipment will need to be picked up from the education office the following morning (Sunday).
Abstracts will not be approved after November 8.Poster Titles and Abstracts
High School Internship in Bioinformatics@PSC
Pallavi Ishwad
High School Internship in Bioinformatics at The Pittsburgh Supercomputing Center, Summer 2010.
A high school research internship program was offered to four students from Pittsburgh area high schools. Two of the students were juniors and two were sophomores. The students selected for this Research Internship have all participated in a pilot bioinformatics course taught from the overall viewpoint of becoming a practicing biologist learning to use bioinformatics techniques and tools in their biology career. The internship provided them with experience in bioinformatics from another angle, that of a computer scientist learning to apply the problem solving and programming skills of his/her discipline to biological sequence data. The goal for this internship project was to examine a set of sequences known to have examples of specific, highly conserved patterns at various positions within the sequences. All patterns are not present in all sequences however. The project aimed to examine sequences with data describing a pattern not known to be present in that sequence according to a standard bioinformatics analysis encoded in a program called MEME. The project attempted to find evidence for a weak, partial, or incomplete presence of the pattern in the sequence. This was intended to provide data as to how sequences with similar but subtly different functions acquire and loose specific patterns of sequence residues allowing that sequence to interact with it or other sequences to carry out specific biological functions. These patterns are associated with inter and intra molecule interactions that define specific biochemical functions and physiological roles of the molecules in which they are found yielding specific details about the acquisition, loss, and modification of these functions and roles within a set of closely related sequences that carry out similar but non-identical functions.
This internship project provided valuable background and experience for these young high school students who has been exposed for the first time to a Bioinformatics curriculum and as a result has become capable to cultivate an interest in a modern scientific career. Based on their learning experience, the two groups of high school students will present Two posters for the 2010 SC Education Program Resource Fair.
Modeling Quantum Electron Transport in Toroidal Carbon Nanotubes with Metallic Leads
Leon William Durivage IV
Carbon nanotubes have been the focus of research due to their many unique physical properties. A carbon nanotube, when wrapped to form a torus, could prove to be a valuable component of nanotechnology. In 2000, a simulation was written in Fortran 77 to simulate electron transport through a (3,3) armchair carbon nanotorus via metallic leads using a recursive Green's function method. Physical properties such as the density of states, the transmission function, and the current can be calculated through this simulation. Over the summer, the code was modified to use the parallel linear algebra package ScaLAPACK and then rewritten in C++ to utilize more efficient data structures which allow for larger simulation sizes, especially on distributed memory architectures. The resulting code is not only more suitable for modern HPC environments, but is also easily modified to include other physical phenomena for future research.
Automated High-Throughput Dereplication and Statistical Analyses of Microbial Diversity Utilizing MALDI-TOF Mass Spectrometry
Benjamin K. Johnson
Scientific computing, and the ability to write specific programs for analysis of large datasets, has become an increasingly important tool in the advancement of biological research. Within the field scientific computing, the use of the programming language Python has become prominent. Python is widely used because it is simple, powerful, and has a large, freely accessible database of scripts written by a community of programmers for use in solving, simulating, and resolving vast amounts of experimental data.
The microbial diversity of an environmental sample is a fundamental question for microbial ecologists. To identify true microbial diversity within a sample, one must distinguish between replicate species and novel species (dereplication) as some species are more numerous and sampled at a higher frequency. Traditional DNA-based methods of dereplication are expensive and analysis can take up to two weeks, whereas protein-based methods such as matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) is capable of rapidly identifying microbes to the subspecies level and can provide results in 48-hours. We propose a novel computational method, written in Python, may be employed specifically for the automated high-throughput dereplication of microbial species using protein spectral output from MALDI-TOF MS analyses. This will provide enhanced capabilities beyond those currently offered by Bruker Flex Analysis©.
We have demonstrated that our novel program contains a set of computational functions that produces similar spectral preprocessing results when compared with the existing analysis program (Bruker Flex Analysis©) with an improvement upon the peak detection algorithm in comparison with Flex Analysis©. Future work will include the addition of statistical analyses when comparing spectra for dereplicative purposes.
Accelerating Geophysics Simulation using CUDA
Brandon Holt
CitcomS, a finite element code for modeling convection in the Earth's mantle, is used by many computational geophysicists to study the Earth's interior. In order to generate more accurate results, finer spatial resolutions are needed, so there is a constant push to increase the performance of the code to allow for more computations to complete in the same amount of time. In order to accomplish this, we leverage the massively parallel capabilities of graphics processors (GPUs), specifically those using nVidia's CUDA framework. We have begun translating existing functions to run in parallel on the GPU, starting with the functions where the most computing time is spent. Running on nVidia Tesla GPUs, initial results show an average speedup of 1.8 that stays constant with increasing problem sizes and scales with increasing numbers of MPI processes. As more of the CitcomS code is successfully translated to CUDA, and as newer general purpose GPU frameworks like Fermi are released, we should continue to see further speedups.
Supercomputing Challenge
David H. Kratzer
Now in its 21st year, the Supercomputing Challenge has been challenging New Mexico students to come up with computational science projects to solve real world problems since 1990. The Challenge promotes computational thinking in science and engineering so that the next generation of high school graduates is better prepared to compete in an information based economy. Middle and high school teams of one to five students and a sponsoring teacher select a project and research it, model it, write about it and present it during the school year long program. Teamwork is a large part of the Challenge.
Time line:
1) Each summer, a two-week Summer Teacher Institute is held to help the teachers become better Challenge team sponsors. 2) In September, teams form and register to participate. 3) In October, a two-day Kickoff Conference is held to excite the students, give them eight hours of classes and have them meet with a scientist to discuss their project proposal, which is posted on the Challenge web site. 4) In December, an Interim report is due which details the progress of the team since the Kickoff Conference. 5) In January, teams are encouraged to make presentations about their project to local audiences. 6) In February, teams make a 30 minute presentation to a small panel of judges at a local university and receive feedback as to how to proceed during the last two months of the Challenge. 7) Typically in March, Challenge teams are offered an opportunity to tour Sandia National Laboratories or the University of New Mexico in Albuquerque. 8) The final written report is due the first Wednesday of April. Teams can also submit a web-based version of their final report for a separate award. A panel of about a dozen finalist judges review all the reports and select about a half dozen teams as first-round finalist. 9) Toward the end of April, all the teams come to the Los Alamos National Laboratory to present their projects. The first-round finalists present to the finalist judges and the rest of the teams present in an Expo. About half a dozen teams are selected from the Expo as second-round finalists and they in turn, present to the finalist judges. 10) After the teams make their presentations, they are taken on tours of the Los Alamos National Laboratory to learn about current research topics. 11) An Awards Ceremony takes place the following day where plaques, cash prizes and scholarships are given out. 12) Typically 300 students on about 80 teams participate in the Challenge. 13) The Supercomputing Challenge is currently sponsored by Los Alamos and Sandia National Laboratories along with support from several universities and many businesses.
More information is available on the Challenge web site: http://www.challenge.nm.org
An eTextBook Providing Blended, Multimodal Access to Computational Physics Curricula
Landau, Rubin H
Modules have been created that encapsulate video recordings of lectures and animated slides covering individual topics from the previously published paper text, "A Survey of Computational Physics, Introductory Computational Science" (Princeton 2008), with the Python version now made into an eBook. These modules have been included into a new electronic textbook integrating the developed video-slide modules, text materials, interactive programs, animations, sounds, and dynamic mathematics (MathML) into a highly accessible form destined for the National Science Digital Library/Compadre, and perfect for tablet computers. The video modules as well as the eTextbook will be demonstrated.
Problems for Introducing Non-Majors to Modeling and Simulation in a Service Course
David Toth
We introduce the idea of modeling and simulation in our service course, Introduction to Information Technology, in order to give non-STEM students an idea of how computing can be useful for solving mathematical and scientific problems. One desired learning outcome of introducing modeling and simulation is that the students will be able to take a new problem and analyze it to determine what they need to know in order to solve the problem. The required knowledge includes how a particular system works as well as the input parameters that affect the behavior of the system and the outcome of the simulation. Students should also learn how to explain how the system works in prose, incorporating the parameters, so that another person can read their description and understand how the system works. The introduction to modeling and simulation also provides the transition in the course to the section on parallel and distributed computing, motivating the need for multi-core computers, clusters, supercomputers, and networking technologies. We present the problems we used to introduce students to modeling and simulation in the course this year which include a coin flipping problem, a dice game, a variant of the game chuck-a-luck, and roulette.
Associate of Science Degree in Computational Science (Biology)
JZorko,ASanders,SGordon
The poster outlines the NSF Grant funded activity which established the Associate of Science Degree in Computational Science (Biology) between the OSC and three community colleges in Ohio. The program goals, competencies, module components and curriculum are highlighted and described.
Where Play and Learning Meet! Using Robotics to Teach Computer Science in Middle and High School Education
Stephen E. Gareau
This presentation will describe a three-week, graduate-level robotics course for technology educators that was offered by the Department of Technology at Buffalo State College, and took place during the summer of 2010. The presentation will describe course objectives, course outcomes, how computer science (particularly C programming language) was integrated into the course, and the highly-engaging, motivational aspects of the course. A main purpose of the presentation will be to examine how robotics instruction can integrate with the teaching of computer science.
Parallelization of Particle-Particle, Particle-Mesh Method within N-Body Simulation
Nick Nocito
The N-Body problem has become an intricate part of the computational sciences, and there has been rise to many methods to solve and approximate the problem. The solution potentially requires on the order of calculations each time step, therefore efficient performance of these N-Body algorithms is very significant [5]. This work describes the parallelization and optimization of the Particle-Particle, Particle-Mesh (P3M) algorithm within GalaxSeeHPC, an open-source N-Body Simulation code. Upon successful profiling, MPI (Message Passing Interface) routines were implemented into the population of the density grid in the P3M method in GalaxSeeHPC. Each problem size recorded different results, and for a problem set dealing with 10,000 celestial bodies, speedups up to 10x were achieved. However, in accordance to Amdahl's Law, maximum speedups for the code should have been closer to 16x. In order to achieve maximum optimization, additional research is needed and parallelization of the Fourier Transform routines could prove to be rewarding. In conclusion, the GalaxSeeHPC Simulation was successfully parallelized and obtained very respectable results, while further optimization remains possible.
Computational Hydrodynamic Modeling of a Glaciated Lake
Russell Manson, Christine Harvey
Abstract
This poster describes a computational model for natural earth systems and its application to modeling seiches in lakes. The model utilizes a computational grid that is spatially unstructured and constructed from connected arbitrarily placed triangles and rectangles. This makes the grid particularly flexible and adaptable to complex geometries and spatially non-uniform system properties. The computational grid and domain specific equations (hydrodynamics) are described and the numerical solution technique is outlined. Finally the model is applied to a real lake geometry based upon a small glaciated lake in the English Lake District. Future development and application of the model is then described.
Keywords unstructured grid; finite volume; seiche; lake; hydrodynamics; numerical testing.
Improving learning of Chemistry concepts using Computational Chemistry and Mathematica tools
Roxanna Delgado, Nanyaly Santiago, Karen Ricardo, Tech Chen, Nitzy Muñoz, Nicole Massanet, Nadja Solís and Carlos Torres
In our laboratory of research in chemical education and computational chemistry at University of Puerto Rico in Río Piedras, we continue developing computational chemistry projects suitable for education at undergraduate level. The research projects studying π-π interactions in dimers of benzene-like compounds containing boron-nitrogen bonds using ab-initio theory are progressing. It has being a very prosperous journey, during these last three years, where students from general chemistry, computational chemistry and physical chemistry courses have been able to experience using resources and programs like Vensim, Gaussian and other computational engines running from WebMO, Chem Sketch, and now Mathematica. All these programs have been used to structure and orbital visualization and to perform calculations of different applications in these courses. In this presentation we are showing a summary of these works emphasizing in the advances of the research work during the last year and in the educational applications already developed using Mathematica for the Physical Chemistry course.
Experiences with BCCD in a Problem-Based Curriculum at Halau Lokahi Charter School
John Rader and Susan T. Brown, Ph.D.
As part of the School and University Partnership for Renewal in Mathematics (SUPER-M) program at University of Hawai'i at Manoa (UHM) Mathematics Department, funded by the National Science Foundation (NSF), a team consisting of a graduate fellow and faculty adviser joined with a high school group at a local charter school, Halau Lokahi, to incorporate High Performance Computing (HPC) into their problem-based learning curriculum. They chose to work on a problem involving climate change, in particular, the effect of sea level change on the coastlines of the Hawaiian Islands. This problem is especially important because of the dramatic effect a relatively small sea level change can have on Hawai'i's economy and population. The students first looked at simple models with Google Earth to see what kind of things to look for, they realized that, for detailed models, they would need HPC. We chose to use the Bootable Cluster CD (BCCD) so the students could also learn about clustering and elements of HPC. BCCD overlays on networked PCs, and is non-destructive, so the PCs can be used when the BCCD cluster is not in use. This was especially important since their math classroom did not have a working computer lab until the project came to the school. To use the BCCD, the students were taught Linux, VI-editor, and simple scripting. We chose to use Micro-DEM as the visualization software, using data from the U.S. Geological Survey (USGS). All the software we used was open source and readily available, and the data was also publicly available. One of the challenges we encountered during the project was the hardware. Computers made available to us were donated, and therefore inconsistent in OS, sometimes locked by password, and not equipped with the latest and best technology. We also did not have internet in the space provided for several months while the school navigated the complex acquisition process. However, once internet access was acquired, we were still able to build a BCCD cluster with the available hardware. The team met with the class twice per week for 50 minutes per class for the majority of the spring semester. During this time, we were able to build the lab, learn Linux, build a BCCD cluster, learn about how climate change affects the planet, how to acquire data from the USGS webpage, and learn how to use the visualization software to view said data. Unfortunately, there wasn't enough time to finish the project. In the spring semester, the school will continue the project with a dedicated instructor and, using the groundwork the team laid in 2010, will port the parallel version onto the cluster and run their models on it.
Bioinformatics in high school
Rebecca Day
Four years ago, several teachers attended a conference to develop a high school level class on Bioinformatics. The program was named B.E.S.T., which stands for Better Educators of Science Tomorrow. The project was ambitious, to say the least. The result of this collaboration with the Pittsburgh Supercomputing Center's Outreach Education Program, met and may have exceeded expectations. Within the 2009/2010 school year, three schools piloted the Bioinformatics class. Each school used the program to varying degrees. Two schools were private and one was a small, rural public school.
The public school Bioinformatics program became part of the Sophomore Honors Biology class. Incorporation of Bioinformatics into the General Biology topics gave students a good perspective on the use of technology within the subject. Although the material was, at times, difficult to teach to students so young, the process proved beneficial to both teacher and student. Many students expressed interest in the field of Bioinformatics as a career. Four students from the three schools participated in a summer research internship at the Pittsburgh Supercomputing Center. Through the internship, the students gained experience in programming using Python as well as VMD and JMol.
The collaboration between public and private schools, along with the Pittsburgh Supercomputing Center has proved fruitful for all involved. The future of Biology and teaching lies in a more integrated and complete understanding of the topics presented using dynamic tools and methodologies. Through the B.E.S.T. program and others like it, innovations in teaching high school students can continue to provide the BEST science education experience.
Heat transfer in nanocomposites by Monte Carlo simulations
Khoa Bui and Dimitrios V. Papavassiliou
Carbon nanotubes (CNTs) and nanographene sheets (GSs) have been suggested to be reinforcement fillers in a variety of composite materials due to their exceptional electrical, thermal and mechanical properties. In terms of thermal properties, incorporating CNTs or GSs into a polymer matrix should increase the effective thermal conductivity of the resulting composite. However, the presence of resistance to the transfer of heat at the nanoinclusion-polymer interface, known as the thermal boundary resistance or Kapitza resistance, results in underperformance of nanocomposites in terms of thermal properties.
Understanding Kapitza resistance and more important, being able to control the heat transfer process in producing superior thermal nanocomposites is a challenging task in experiment due to the difficult in precisely controlling the dispersion pattern and measuring Kapitza resistance. By treating the heat flux as flow of discrete heat walkers, we apply Monte Carlo simulation to calculate the effective thermal conductivity of CNT nanocomposites taking into account the Kapitza resistance, as well as different inclusion geometries (sphere, cylinder and parallelepiped). The effect of the dispersion pattern of the nano-inclusions is also investigated. Finally, comparing the calculated thermal conductivity from the simulations to experiments, the methodology can be used to calculate the Kapitza resistance of such systems.
Blue Waters Undergraduate Petascale Education Program
Jeff Krause, Bob Panoff, Scott Lathrop
The Blue Waters project aims to revolutionize computational science and engineering education by promoting new educational resources, models and methods that transcend traditional boundaries of discipline and institution. Over the next three years, through a systematic effort of materials development, faculty workshops, and student internships, we will greatly increase the nation's capabilities to address the educational opportunities offered by petascale computing across all undergraduate fields and will broaden participation in high performance computing. Faculty from 2 year and 4 year colleges and universities are invited to contribute to the development of inter- and multi-disciplinary modules demonstrating modeling approaches and applications at the petascale. Five week-long faculty workshops a year will provide training for faculty in taking their scientific modeling to the petascale. Undergraduate internships will provide up to a year of funding for rising juniors and seniors who are interested in working with a petascale research group.
Discovering Data In The Classroom: The SDSC Discover Data Portal
Jeff Sale, Ange Mason, Diane Baxter
For over twenty years, the SDSC Education Group has offered support to K-12 educators through workshops and seminars that provided teachers access to the latest developments within a wide range of science and engineering fields, as well as the latest tools and technologies to use in their classrooms, including topics such as web development, Smartboards, podcasting. Two years ago we decided to take our first major step towards supporting online synchronous and asynchronous learning by providing teachers in San Diego County who have participated in our workshops and seminars free accounts on the TeacherTECH Community Portal (TTCP). The TTCP uses the open source Moodle course management system. Teachers attend workshops on how to use Moodle, and they use the TTCP as they would any learning management system, i.e. to organize and manage their students learning experience. This paper documents the benefits offered by the TTCP and the D2P, and discusses some of the challenges we have faced in their establishment as well as solutions either implemented or planned to address these challenges.
Quantum Educational Mechanics
Byron Sharer Robertson
The Achievement Gap represents one of the most formidable challenges to industry leaders, researchers and educational practitioners. The achievement gap is significant because it represents a 2-3 year academic gap between Anglo-American and Asian students on the one hand, and African, Hispanic and Native-American students on the other. In a time where record numbers of individuals will be retiring from science and math intense fields and the minority population in the United States will be growing, the need for knowledge workers and leaders in the areas of science and math is at an all time high. While traditional educational reforms have been attempting to close this achievement gaps for the past 60 years, the results have been mixed and the gap remains. This research proposal hypothesizes that a non-traditional modified community involvement intervention (with high school minority student participants developing a supercomputer and using its science and math applications to redevelop their own community) may produce a quantum leap (Cohen's .80) in student achievement. The community involvement intervention is informed by the Learning First and the PPSI framework that examines how educational organizations can partner with community institutions, organizations and assets to produce a dramatic increase in student achievement. Specifically, the research will utilize a random control trial that will test whether the community involvement intervention has the statistical and dramatic effect that is hypothesized.
Beyond Astro 101 -- Examining the Lower-Division Astronomy Curriculum for the 21st Century
Lancelot L. Kao, Orkan M. Umurhan, and Theresa J. Summer
So-called "ASTRO 101" survey courses in general astronomy are offered to non-science majors in colleges and universities across the United States, to fulfill general-education requirements in the physical sciences. At least two of the common Student Learning Outcomes (SLOs) for these courses are critical thinking and understanding astronomy as a scientific discipline. We argue that a comprehensive lower-division astronomy program surpassing ASTRO 101 would increase science literacy for non-science majors, STEM students, and the general public. The program would include diverse astronomy course offerings, interdisciplinary science courses (e.g. astrobiology), service-learning and peer-mentoring activities, and internship opportunities.
Visualizing Chemical Principles Using High-Quality Calculations
Jay S. Shore
Many people consider themselves visual learners and educational research has indicated that visualizations can lead to an increase in conceptual understanding. The main goal of my research in visualization of chemical principles is to understand what makes a visualization effective. To determine this, I am trying to answer the following questions: what types of visualizations are the most effective (e.g., is a Java applet that students can manipulate more effective than a movie that they just watch?); can students be motivated to spend more time studying by making high quality visualizations accessible?; How affective is a visualization shown in class, versus one shown in class and incorporated in an online quiz?; Are visualizations more effective when they are spread throughout the curriculum.
During the past few years as part of my research, I have produced visualizations (of arbitrarily high resolution in a variety of formats) from calculations of relatively high quality. I believe that the calculations have to be of relatively high quality, otherwise there may be inaccuracies in the visualization that lead students to an incorrect conclusions. For example, when the potential energy of two fluorine atoms is calculated as a function of distance at lower levels of theory incorrect conclusions are often made. While the CCSD(T) calculation is fairly accurate, the SCF calculation yields an inaccurate bond distance and does not include the increase in potential energy as the atoms approach each other. The calculation done using second-order perturbation theory (MP2) becomes very unrealistic as the atoms become further than 2 Angstroms apart. To accurately visualize the potential energy as a function of bond distance or bond angle, requires a fairly high level of theory.
Visualizations illustrating different chemical principles, created from relatively high-quality calculations will be presented.
Institute for Chemistry Literacy through Computational Science
Kathy Hughes, Lori Reed, Mike Wallace, Dave Mattson, Edee Wiziecki
The Institute for Chemistry Literacy through Computational Science (ICLCS) is a program of the University of Illinois' Department of Chemistry, College of Medicine, and the National Center for Supercomputing Applications, K-12 partners A-C Central School District, Illinois Regional Office of Education #11, and 118 additional school districts across Illinois representing 124 ICLCS Fellows. This program is a 5-year National Science Foundation, Math Science Partnership program to increase the chemistry literacy of teachers and their students and to integrate computational science into the high school curriculum of rural Illinois schools.
Multi-Channel Radar Depth Sounder (MCRDS) Signal Processing: A Distributed Computing Approach
Jeaime Powell, Linda Hayden
In response to problems surrounding measuring ice sheet thickness in high attenuation areas of the Arctic and the Antarctic, the Center for the Remote Sensing of Ice Sheets (CReSIS) from the University of Kansas created a Multi-Channel RADAR Depth Sounder (MCRDS). The MCRDS system while tuned between the VHF 50-250MHz ranges was used to measure ice thicknesses of up to five kilometers in depth. While receiving measurements, large datasets were generated producing a signal-processing bottleneck. The purpose of this project was to test processing performance increases on a 32-core cluster through distributed computing resources. Testing involved a six-node cluster with an attached storage array and use of the CReSIS Synthetic Aperture RADAR Processor (CSARP) distributed through the MathWorks MATLAB Distributed Server. Performance testing stemmed from average run times collected once CSARP jobs completed. The run times were then compared statistically using an ANOVA test proving performance benefits in job distribution.
Computational Water Resources at the Northern Gulf Institute, Mississippi State University
Vladimir J. Alarcon, William H. McAnally, James L. Martin ,Yi Xiong
The modeling and simulation of water resources processes is changing rapidly. Recent developments in information technology, remote sensing, and data collection technologies have changed the paradigms under which modeling is done. This change requires that water resources modelers find novel approaches from conceptual models and numerical techniques, up to results visualization and dissemination of results. This poster presents an up-to-date account of current modeling techniques, simulation/visualization strategies, and results of several on-going projects at the Northern Gulf Institute, Mississippi State University.
Back to the Future or Forward to the Past?
Marian Houseman, Bruce Bennett, Rosalie Eimers
Presented by the North Polk Jr/Sr High Team from Alleman, Iowa: Bruce Bennett => technology coordinator/computer science; Rosalie Eimers => chemistry; Marian Houseman => 7-12 grade talented and gifted consultant. Information will be shared on past and current uses of computational science in grades 7-12 at North Polk Jr/Sr High.
North Polk has had a team of teachers involved in computational science since the SC00 conference. Their efforts have included integrating STELLA, Vensim, Excel, SketchUp, and Jeroo into various curricular areas. North Polk's introductory computer programming course changed to an introductory computational science class which incorporates activities from Shodor's Interactivate site, Excel spreadsheets and computer programming. Team members have attended summer NCIS workshops on Introduction to Computational Science and curriculum writing as a follow up to attendance at several SC Education Program conferences. The team is currently working on developing a new high school class for 2011-12 which will involve students working in teams to solve self-selected real life problems using computational science.
Performance Analysis of Cluster Computing
Briana Johnson, Gabrielle Buchanan, Stephen V. Providence
Message-Passing Programming is foundational to our work. Most parallel programming continues to be done in either Fortran or C augmented with functions that perform message-passing between processes. The MPI standard is the most popular message-passing specification supporting parallel programming. In the future we will use the following the basic functions to do simple parallel programming as follows: MPI_Init, initialize MPI; MPI_Comm_rank, processes' ID number; MPI_Comm_size, number of processes; MPI-Reduce, reduction operation; MPI_Finalize, to shut down MPI; MPI_Barrier, barrier synchronization; MPI_Wtime, to determine the wall time; MPI_Wtick, to find timer accuracy. Benchmarking is our next step. We compute: χ = time for single iteration of loop to complete inner product; χ [n^2] = expected time for the computational portion of parallel program. We want to determine the values of λ and χ to determine how well the parallel programming actually affects the performance of the cluster computer. One application is Matrix-Vector Multiplication using Cauchy-like and Vandermonde-like matrices.
As an example, we determine output values of k-level neural network input values using n x n matrix by vector multiplication. We will use several MPI programs, each will be based on different data (SIMD model). Issues are different matrix and vector elements; computing complexity of Matrix-Vector multiplication, and use of n-element vectors. This requires n multiplications and n-1 additions where complexity of inner product of two vectors: θ(n); complexity of Matrix-vector multiplication: θ(mn); complexity of a square matrix by vector: θ(n^2) and finally parallel complexity is proportional to: θ(χ n2/p). We intend to report the results of our analysis in the near term.
Parallel Suffix Tree Construction For Next Generation Genomic Sequences
Hao Lu
Suffix trees are a well know data structure for indexing strings and sequences to support efficient pattern matching. In the field of computational biology, sequence data are accumulating exponentially, hence there is a critical need for faster and scalable suffix tree construction algorithms. In this work we propose a new parallel algorithm for constructing suffix trees over short-reads generated from next generation sequencing methods. Our approach is designed for the distributed memory machine model. Our algorithm integrates the DC3 linear suffix array construction method into the parallel construction process. The algorithm runs in O(nl/p)+pt +m(n) time and has O(n/p) memory complexity, where n is the input-size, p is the number of processors l is the average length of sequence, t is the connection overhead, and m is the transfer rate per- byte. We are currently implementing this algorithm in C++ / MPI.
3D Parallel Implementation of Finite Element Method with PETSc
Benjamin Cousins
We study a parallel implementation of a finite element method for 3D Stokes/Navier-Stokes equations of incompressible Newtonian flow, using the PETSc library for parallel scientific computing. We find a nearly linear scaling of processors to computation time for the assembly part of the code (mesh generation, matrix assembly). For the linear solve, we find the best method (in terms of accuracy and robustness, of the built in linear solver of PETSc) was block-Jacobi preconditioned bicgstab; this does not scale linearly with the number of processors, but parallelization still significantly improves efficiency. Tests were performed on a model problem (up to ~750,000 DOF) and for 3D channel flow over a forward-backward step. For future work, we plan to improve the scheme by implementing enhanced physics based methods for the Navier-Stokes equations and by using higher order approximation elements.
Teaching with ESTEEM
Srebrenka Robic
How can we engage our students in explorative learning with real data? Various online databases contain abundant sequence data sets, which can serve as an exciting starting point for student research. The BioQUEST Biological ESTEEM (Excel Simulations and Tools for Exploratory Experiential Mathematics, http://bioquest.org/esteem/) is a collection of Microsoft Excel-based modules exploring concepts bridging mathematics and biology. One of these modules, Protein Analysis, can be used in introducing students to exploration of protein sequences. Students input a protein sequence, and explore various parameters such as amino acid and charge distributions through various visualization tools. Suggestions for incorporating Protein Analysis into undergraduate teaching will be discussed.
Drug Discovery in a Cloud
Sally R. Ellingson and Jerome Baudry
Discovering new drugs is an expensive and lengthy process with a low-yield of new drugs available on the market. Virtual docking is a tool developed to screen databases of chemical compounds to find compounds with favorable binding properties to a target protein, therefore reducing the cost and discovery time for novel drugs. Large scale virtual docking requires considerable computational hardware and expertise, limiting the use of this powerful technology to computational laboratories that posses such computational wealth and experience. Classrooms, small labs, and other groups need a computational technology that leverages the power of existing docking engines without requiring a large investment in hardware or expertise. The aim of this research is to leverage the power of cloud computing to make high-throughput virtual docking available “to the masses” and to the drug discovery laboratories. The research will bring a modern and efficient computational technique to the desk of scientists and to the classroom. We anticipate a large amount of usage from the scientific community, and that such research will greatly increase the discovery of new compounds binding in newly characterized proteins, contributing to fulfilling “post-genomics” goals of synthesizing knowledge from the massive amount of available biological data.
Keywords: cloud computing, high-throughput virtual docking, drug discovery
Blue Waters: Towards petascale sediment transport computations in rivers
Laielli, M; Page, R; Manson, JR
The project is one of several undergraduate Blue Waters projects. In this project the students have developed and begun testing HPC codes for predicting sediment transport in rivers. The model is based on a particles tracking approach with billions of particles involved. Particles are advected with the mean river velocities and buffeted by turbulent eddies; currently there is no sub-model for particle-particle interaction however that is planned for the next phase of model development. The HPC code utilizes MPI to access multiple cores to achieve faster and bigger computation and current results and speed-ups are shown. Future coding efforts will focus on using CUDA to allow each of the multi-cores to access a many core GPU. In this way we believe the computation can be taken to the petascale. The faculty advisor is working on an undergraduate lesson plan which will utilize this code to teach sediment transport physics and multi-core/many core computational paradigms.
Building Content Communities To Support Computational Science Education
Jennifer Houchins, Patricia Jacobs, Robert M. Panoff
Shodor and its partners in the Computational Science Education Reference Desk (CSERD) are "transforming learning through computational thinking." CSERD has engaged key partners and collaborators to leverage its computational science resources and to build content communities in computational science education. This poster showcases such key partnerships as HPC University, Blue Waters and HiPOP that allow CSERD to serve a wider community of educators and students and to foster communities of learners in High Performance Computing. In addition, CSERD continues to enhance its own collection by the addition of portal projects for the sub-domains of computational science such as Computational Chemistry for Chemistry Educators (CCCE), Computational Biology for Biology Educators (CBBE) and Computational Physics for Physics Educators (CPPE). This poster also highlights CSERD's upcoming Journal of Computational Science Education (JOCSE). JOCSE promotes the use of computation in education through disseminating unique uses of computation in the classroom as well as research findings in computational science education, with submissions from both professionals and students. JOCSE utilizes Internet technology and a web-based format to allow for enhanced interactivity in all publications. Materials accepted by JOCSE will be hosted on the JOCSE website, and will be catalogued by CSERD for inclusion in the National Science Digital Library.
Modeling the Behavior of the Ceph Parallel File System for Detecting and Classifying Performance Faults
Steena Monteiro, Marc Casas-Guix, Greg Bronevetsky
Today's largest applications are characterized by high parallelism and extreme complexity. The high number of components contained in these applications and their cohesive interactions provide a rich domain of complex system behaviors. In order to provide an effective fault detection and classification scheme for these applications, we investigate and profile the working behavior of Ceph―a scalable, high performance parallel file system. Using Ceph as the current experimental artifact, the research goal is to ultimately develop automated techniques to identify context information on which the application’s behavior depends using combinatorial algorithms and statistical learning classifiers. Current results show that Ceph's system initialization workings can be distinguished from its steady state operations. Further, results from profiling different message processing functions in Ceph's Object Storage Devices (OSDs) reveal that different workloads exhibit predictably varying profiles. Future work includes using the collected contextual information to train a statistical model, which will provide an insightful representation of complex application behavior. The results from the model will be used to fine tune the application and detect failures.
Navigating Protein Design through Computational Biology
Yolanda Hom, Daniel Hsieh, Kenneth McGuinness, Douglas Pike, Vikas Nanda
The Nanda Lab is an interdisciplinary biological research group that studies an array of biochemical and biophysical questions regarding protein evolution, structure and function.
We are a diverse team that utilizes both advanced computational and experimental tools to further medical and biological research.Our studies employ bioinformatic techniques to explore protein evolution via instrinsic codon table selection and gastric digestion of food allergen proteins. We also design and model the insertionability of transmembrane beta-barrel proteins and the self-assembly of heterotrimeric collagen-based biomaterial. We take a reductionist approach to solving these biological questions, which allows for a finer analysis of isolated parameters. However, our current computational power and infrastructure limits search space, resolution of our simulations and restricts our choice of parameters; preventing us from answering more complex and important biological questions.Therefore we see HPC as a new toolbox that would advance our research in the complex and ever-changing biological environment.
