INSTITUTE FOR

DATA AND HIGH PERFORMANCE COMPUTING[ ]Leadership in Scientific Discovery

FROM THE DIRECTOR[ ]Data intensive and high performance computing (HPC) are playing essential roles in attacking the most important problems that face society today. The Institute for Data and HPC (IDH) provides a fertile environment and an organizational framework to create and advance innovative, multidisciplinary research efforts in these areas. Georgia Tech faculty make up a robust community of researchers spanning a broad range of key areas including applications, algorithms, hardware and software systems, and the underlying foundational mathematics. We are proud of Georgia Tech’s reputation and track record for multidisciplinary research and our ability to develop and apply technology toward the solution of science and engineering’s biggest problems. I invite you to learn more about our talented researchers and exciting research and education programs by reviewing this brochure and visiting our website at www.idh.gatech.edu. Please don’t hesitate to contact us for additional information. Sincerely,Richard FujimotoInterim Director, Institute for Data and HPCRegents’ Professor and Chair, School of Computational Science & EngineeringGeorgia Institute of Technology

Leadership in Scientific Discovery 1

High performance computing is about much more than speed—it’s about what you can achieve with extreme capability. At Georgia Tech, we don’t just have people and resources; we have the right people and the right resources to lead computation-driven scientific discovery. From experimental systems to applied research, we are addressing the world’s most important challenges, including:

Healthcare — diagnosis of heart disease and stroke, identification of tumors, detection and prevention of epidemics and pandemics

Bioinformatics and Systems Biology — protein folding, drug design, biological system simulation, complex life systems, microbial research, HIV virus, life and evolution

Sustainability and Urban Infrastructure — city planning, intelligent transportation systems, communication, water, food supply, emergency planning

Energy — smart electric power grids, combustion, fusion, energy conservation behavior

Nanoscience and nanotechnology — nanomedicine, manufacturing, electronics and supercomputers, consumer products

Cybersecurity — Web science and interaction networks, Internet security, social networks

Data-Intensive Analytics — business analytics, streaming graph problems, national security

Homeland Security — text analysis, fingerprint and face recognition, data and visual analytics

IDH leverages investments and research in data intensive and high performance computing at Georgia Tech for the benefit of our modern society. We do this by creating and strengthening multidisciplinary research teams that combine deep knowledge of HPC application areas with advances in computational techniques and foundational mathematics to attack the most challenging problems facing society in science, engineering, and the social sciences. New innovations in computational methods must be transitioned to useable tools and software to advance research in the application domain. Traditional computational research often stops short of creating such codes once a research prototype has been created. Computational artifacts, which provide tangible value to researchers and can be exported beyond Georgia Tech, offer a critical avenue for increased impact of research innovations. Further, such codes are essential to explore the realization of next-generation HPC machine architectures and systems.

Suresh Menon, professor in the School of Aerospace Engineering, uses large eddy simulation techniques to capture the physics of swirling spray flames in gas turbine combustors in order to predict how droplet dispersion, fuel-air mixing, and flame stabilization can impact overall emissions.

In this snapshot from a dynamic simulation of the macromolecular environment inside a cell, different colors represent macromolecules of different sizes. (Image courtesy of Tadashi Ando and Jeffrey Skolnick)

2 Georgia Tech [Institute for Data and High Performance Computing]

DATABig

Navigating a Sea of DataAs recently as a decade ago, the challenge in data analytics was in gathering adequate amounts of data. Today the challenge lies in making sense of the oceans of data that are now available. Haesun Park is tackling that challenge in her role as the principal investigator of the Georgia Tech team leading the Foundations of Data and Visual Analytics (FODAVA) program, sponsored by the National Science Foundation and the U.S. Department of Homeland Security. In its role as the lead institution of the FODAVA program, Georgia Tech is working with nineteen other universities to develop the data and visual analytics field. Data is everywhere. Some of the largest data sets these days are used in areas such as network security, healthcare, bioinformatics, and homeland security. The FODAVA program aims to capitalize on knowledge and expertise in mathematics, computational science, information visualization, and cognitive science to produce new methods to “detect the expected and discover the unexpected in massive data sets.” The problem is complicated further by the fact that some data doesn’t lend itself to analysis using numerical methods, Park noted. In many fields, tens of thousands of free-form, unstructured documents—such as e-mail exchanges and doctors’ notes—are being collected, and people need a way to systematically analyze them. She is hoping to help find the way. “The kind of work I did as a pure mathematician felt too theoretical, too isolated from real life,” said Park. “The work I do now provides very foundational understanding of problems. But at the same time, I try to work very closely with the people in the application domain. Theory provides a foundation for applications, and applications provide important insights into theoretical work.” “I have worked on some problems where everybody thinks the ultimate solution has already been found,” said Park. “But sometimes I am able to go back and approach it differently and find an even better way to do it. It’s really nice when that happens.”

Professor and Associate Chair, School of Computational Science & Engineering

PhD, Cornell University

Professor and Executive Director of High Performance Computing, School of Computational Science & EngineeringPhD, University of Maryland

Center for Adaptive Supercomputing Software for Multithreaded ArchitecturesThe newest breed of supercomputers have hardware set up not just for speed, but also to better tackle large networks of seemingly random data. A multi-institutional group of researchers will develop software for these supercomputers. The software will be useful anywhere complex webs of information can be found: from Internet security and power grid stability to complex biological networks. The Center for Adaptive Supercomputing Software for Multithreaded Architectures (CASS-MT) is a national center in partnership with Pacific Northwest National Laboratory with technical leadership from Georgia Tech. Led by David Bader, Georgia Tech’s team directs efforts within the CASS-MT to develop methods for analyzing massive complex semantic networks. More thoroughly understanding social networks, for example, can help tackle challenges such as influencing change and understanding critical trends in behaviors and customs. Similar computational capabilities can be applied to finding vulnerabilities in the power grid and monitoring important protein interactions in cancer research.

P R O F I L EHaesun Park

P R O F I L EDavid Bader

Leadership in Scientific Discovery 3

KEENELAND CUDAand

The Keeneland ProjectIn 2009, the National Science Foundation (NSF) selected a team led by Georgia Tech to develop and deploy a large scale heterogeneous

supercomputer for open computational science. This innovative system uses Fermi graphics processors from NVIDIA as accelerators. These accelerators provide very high memory bandwidth and fine grained parallelism when compared to commodity microprocessors. Led by Jeffrey Vetter, joint professor of the School of Computational Science and Engineering at Georgia Tech and Oak

Ridge National Laboratory, Keeneland operates in partnership with the University of Tennessee-Knoxville and Oak Ridge National Laboratory. Keeneland will initially acquire and deploy a small, experimental, high performance computing system consisting of a system from Hewlett-Packard with attached NVIDIA accelerators in 2010. In 2012, the project will upgrade the heterogeneous system to a larger and more powerful system based on a next-generation platform and NVIDIA accelerators. It is anticipated that the final system will have a peak performance of approximately two petaflops. Keeneland will be integrated into the TeraGrid to make the system more widely accessible to the research community. The project team will use this system to develop scientific libraries and programming tools to facilitate the development of science and engineering research applications. The project team will also provide consulting support to researchers who wish to develop applications for the system using OpenCL or to port applications to the system. The final system has the potential to support many different science areas. Possible areas of impact include some of the scientific domains in which GPU-based acceleration has already been demonstrated to have an impact at smaller scale, for example chemistry and biochemistry, materials science, atmospheric science, and combustion science. In addition to providing infrastructure for science and engineering research and education, the project partners will educate and train the next-generation of computational scientists on cutting-edge computing architectures and emerging programming environments, using the experimental computing resource as one example.

CUDA Center of ExcellenceIn 2010, NVIDIA designated Georgia Tech as a CUDA Center of Excellence. Georgia Tech is among ten other universities and research organizations in the United States and abroad that were given this distinction. Jeffrey Vetter serves as principal investigator of the CUDA Center of Excellence. “Georgia Tech has a long history of education and research that depends heavily on the parallel processing capabilities that NVIDIA has introduced with its CUDA architecture,” Vetter said. “This award allows us to focus what is now a large amount of activity across twenty-five different research groups, under a single center, which will significantly amplify our research capabilities.” CUDA is NVIDIA’s parallel computing architecture that enables dramatic increases in computing performance by harnessing the power of a GPU. Georgia Tech is engaged in a number of research, development, and educational activities that leverage GPU computing. These activities span the full gamut: applications, software development tools, system software, and architectures. Partners include Georgia Tech, Georgia Tech Research Institute (GTRI), Centers for Disease Control and Prevention (CDC), Oak Ridge National Laboratory, and Accelereyes.

PRO J EC TS

4 Georgia Tech [Institute for Data and High Performance Computing]

Algorithmic Health: Battling Disease with ComputationAccording to World Health Organization estimates, 17 million people around the world die of cardiovascular diseases each year. George Biros is working on supercomputing applications to tackle this major health problem. Strokes, heart attacks, pulmonary embolisms, and other conditions can be caused by thrombosis—blood clots that form in veins or arteries and can travel, sometimes with fatal results, to the brain, heart, or lungs. Biros, associate professor in the School of Computational Science and Engineering with a joint appointment to the Coulter Department of Biomedical Engineering, uses high performance computing to study several aspects of the cardiovascular system, from flow in small capillaries to the mechanical properties of ischemic myocardium, in hopes of finding a way to counter thrombosis and its deadly effects. “We want to answer questions about the causes of thrombosis,” Biros said of his current research, “because that is important if you want to design new drugs for thrombosis or if you want to design stents for bypass surgery or new mechanical heart valves that won’t cause clots to form.” Experimentation on living patients is nearly impossible, so researchers like Biros have turned to in vitro experiments and computer modeling and simulation for answers. One basic problem with computer simulations concerns understanding the hydrodynamic interactions between red blood cells and platelets. The number of blood cells that affect blood flow and clotting is so enormous that ordinary computers cannot perform the necessary calculations with accuracy and speed. That’s where supercomputing comes in. “About four years ago, we developed these algorithms that allowed us to speed up calculations by five orders of magnitude and allowed simulations of hundreds of thousands of cells—calculations that previously had taken one year to complete could be done in one day or less,” Biros said. “This tool has enabled new discoveries, and I consider it to be the part of my work that has had the biggest impact.”

Astrophysics and High Performance ComputingAs a graduate student, Pablo Laguna was interested in mathematical relativity and quantum gravity. He then discovered that his real passion was in numerical relativity, using the power of supercomputers to explore black holes, gravitational waves, and neutron stars, all studied through equations governed by Einstein’s Theory of General Relativity. As a postdoctoral fellow at Los Alamos National Lab and later as a faculty member at Penn State, Laguna moved his research into the interface of gravitational physics and astronomy. Supercomputing has dramatically changed the landscape of general relativity, and numerical relativity—the formulation of Einstein field equations in a way amenable to numerical analysis—has emerged as a field of its own. Because of the complexity of the Einstein equations, very few exact solutions are known, and numerical techniques are often the only avenue for exploring general relativity. “To learn astronomical phenomena that shape our universe, traditional astronomy uses electromagnetic observations,” said Laguna, professor of physics and director of the Center for Relative Astrophysics. “But there are

objects, such as black holes, that are intrinsically dark or phenomena obscured by dust and gas for which other means of observing is needed. Neutrinos, cosmic rays, and gravitational waves provide those new avenues.” The motivation behind his work is the development of computational tools and methodologies that help analyze the data collected by gravitational wave interferometric detectors such as LIGO (Laser Interferometer Gravitational-Wave Observatory). Laguna and his team are engaged in multimessenger research involving electromagnetic, gravitational, and particle astrophysics. “[Only recently has this] field of gravitational physics become observationally driven.” Laguna said. “When I started, I could not have imagined the possibility of observations directly connected to this work. Observations of the gravitational waves emitted by black hole and neutron star binary systems are just around the corner. These observations will open a new window to the universe and test Einstein’s theory of general relativity in the most extreme situations.”

APPSPeta

A picture from a numerical simulation of the collision of two supermassive black holes in the center of a galaxy. (P. Laguna, T. Bode, T. Bogdanivic, R. Haas, D. Shoemaker)

Right: George Biros

Leadership in Scientific Discovery 5

Understanding Genomic Evolution with Petascale Computational ToolsTechnological advances in high-throughput DNA sequencing have opened up the possibility of determining how living things are related by analyzing how their genes have been rearranged on chromosomes. However, inferring such evolutionary relationships in this way is computationally intensive, even on the most advanced computing systems. “Genome sequences are now available for many organisms, but making biological sense of the genomic data requires high-performance computing methods and an evolutionary perspective, whether you are trying to understand how genes of new functions arise, why genes are organized as they are in chromosomes, or why these arrangements are subject to change,” said David Bader, professor and executive director of High Performance Computing in the School of Computational Science and Engineering. Even on today’s fastest parallel computers, it could take centuries to analyze genome rearrangements for large, complex organisms. Thus, the research team is focusing on future generations of petascale machines, which will be able to process more than a thousand trillion calculations per second, compared to a few hundred thousand per second on the average personal computer. The researchers, led by Bader, plan to develop new algorithms in an open-source software framework, using parallel, petascale computing platforms to infer ancestral rearrangement events. On a dataset of a dozen bellflower genomes, the software determined the flowers’ evolutionary relatedness one billion times faster than the original implementation. The next test will analyze a collection of fruit fly genomes, providing a relatively simple system to understand the mechanisms that underlie gene order diversity, which can later be extended to more complex mammalian genomes, such as primates. The researchers believe these new algorithms will make genome rearrangement analysis more reliable and efficient, while potentially revealing new evolutionary patterns. In addition, the algorithms will enable a better understanding of the mechanisms and rate of gene rearrangements in genomes, and the importance of the rearrangements in shaping the organization of genes within the genome. “Ultimately this information can be used to identify microorganisms, develop better vaccines, and help researchers better understand the dynamics of microbial communities and biochemical pathways,” added Bader.

Predicting TurbulenceTurbulence, a subject of great complexity, plays a central role in diverse fields of science and engineering, including aeronautics, astrophysics, meteorology, oceanography, propulsion, pollutant transport, and many others. Our ability to predict natural phenomena (such as clouds in the atmosphere) and to design improved engineering devices (such as more efficient and cleaner combustion equipment) depends in good measure on an understanding of flow physics. Unfortunately, the current understanding is inadequate, especially considering aspects of flow structure not readily amenable to laboratory measurement or theoretical description. Advances in fundamental understanding are expected to have a wide impact on both science and society. With support from the National Science Foundation (NSF) in fluid dynamics and cyberinfrastructure, P. K. Yeung, professor in the Guggenheim School of Aerospace Engineering and adjunct professor in the School of Computational Science and Engineering, is leading a team of researchers in both turbulence and high performance computing to simulate the finest details of turbulence flow structure, as well as turbulent mixing and dispersion at a state-of-the-art resolution of 64 billion grid points. These simulations provide valuable information concerning how an infinitesimal piece of fluid can be severely distorted by turbulence at a high Reynolds number. Furthermore, they allow us to understand the intricate coupling between turbulent transport and molecular diffusion acting at the smallest cases, as well as the motion of large numbers of small particles in relation to the local spatial structure of the flow. The nature of turbulence involving fluctuations over a wide and continuous range of interacting scales also makes the use of high grid resolution very important. A series of production computations with up to thirty-two thousand CPU cores has been performed using large resource allocations provided by several major national centers (Texas Advanced Computing Center, National Institute of Computational Sciences, and National Center for Computational Sciences). Tests at even larger problem sizes have been performed as well. A current focus of code development is to strive toward sustained Petaflop performance on future architectures at the level of the NSF-supported Blue Waters Petascale computing facility.

Visualization of turbulence structure, by staff at the Texas Advanced Computing Center, using red isocontours and blue volume rendering to compare different types of high-activity zones.

6 Georgia Tech [Institute for Data and High Performance Computing]

Nanotechnology Pioneer: Small is DifferentTheoretical physicist Uzi Landman has been a pioneer in the quest to discover how materials behave on the nanoscale and in using computer simulations to discover new phenomena on the nanoscale. In 1999 his team discovered that gold is a very effective catalyst when aggregated in clusters of eight to two dozen atoms and that electrical charging of gold is crucial to its catalytic capabilities. These theoretical predictions have been verified experimentally. “This collaboration, where theory and experiment complement and challenge each other, had already resulted in several key discoveries,” said Landman. “We expect that the continuation and strengthening of the interaction between our research groups, enabled by the Humboldt Award, would open new research directions in nanocatalysis, including in areas related to energy research and environmental issues.” During the past decade, Landman and his coworkers have investigated the properties of electrons confined in quantum dots

fabricated at the interfaces of semiconductor heterostructures and studied as potential logic gates in quantum computers. As early as 1999, Landman, in collaboration with Senior Research Scientist Constantine Yannouleas, discovered formation of crystalline patterns of the confined electrons, called “electron molecules,” which were experimentally verified in a joint project with a group at the ETH in Switzerland and published in 2006. This research has been extended recently to investigations addressing formation of “boson molecules” in ultra-cold trapped atomic systems. “Small is different,” said Landman. “We cannot use the way physical systems behave on the large scale to predict what will happen when we go to levels only a few atoms in size. But we know the rules of physics, and we can use them to create model environments in which we can discover new phenomena through high-level computer-based simulations, which serve as a ‘computational microscopy’, supplementing, complementing, challenging and motivating laboratory experiments. In this way we employ computers and novel computational methodologies as tools of discovery.”

Improving Drug DiscoveryModeling and simulation have played relatively minor roles in pharmaceutical research and development to date. Each drug and target combination is typically considered in isolation, which can often be misleading because the mechanisms contributing to the development of disease are complex and not just the result of the contribution of a single gene or its protein product. With the increased use of antibiotics to treat bacterial infections, pathogenic strains have acquired antibiotic resistance, prompting extensive effort in the design of new or improved antibacterial agents. One target of antibiotics is the ribosome—the cellular

workhorse that translates the genetic code into proteins. To discover where new drugs will bind to the ribosome, researchers must know the structural shape of the ribosome, which includes two subunits that assemble to produce a functional particle at the beginning of the process of protein biosynthesis. Jeffrey Skolnick has worked with postdoctoral scientist Michal Brylinski to develop a method for predicting where signal-triggering molecules called ligands will bond to target proteins. The prediction can be made without knowing the protein structure. The computations required for the predictions involve rating the quality of the fit to various sites on the protein, analyzing the molecules’ abilities to either enhance or disable the function of the protein, depending on its function in the cell. Identifying the ligand-binding site is often the starting point for protein function determination and drug discovery. “It’s much easier to feed cells a molecule that’s been inhibited because of a disease rather than designing a drug to inhibit a molecule that’s being produced in excess due to a disease,” explained Skolnick.

P R O F I L EUzi LandmanRegents’ and Institute Professor, Callaway Chair in Physics, and Director of the Center for Computational Materials SciencePhD, Technion - Israel Institute of Technology

P R O F I L EJeffrey SkolnickProfessor and Director of the Center for

the Study of Systems BiologyGeorgia Research Alliance Eminent

Scholar in Computational Systems Biology

PhD, Yale University

Leadership in Scientific Discovery 7

Ubiquitous High Performance Computing (UHPC), a program of the Defense Advanced Research Projects Agency (DARPA), will develop next generation supercomputing technology capable of delivering more than a petaflop of computing capability for less than 57 KW in a single rack cabinet form factor. The program’s goal is to create a new class of “exascale” computers, at scales ranging from embedded machines to data centers, that will be 1,000 times more powerful than the fastest computers available today. The goal was derived from the findings of several studies conducted from 2007 to 2009 that concluded major leaps in energy efficiency, dependability, and programmability would be required to enable exascale computers by the end of the 2010s. These studies also showed that much greater degrees of parallelism than seen even in current petascale machines would be required to meet the performance and energy requirements, and that fundamentally new execution models must be developed to achieve workable hardware and software designs.

CHASMCHASM, led by the Georgia Tech Research Institute’s Dan Campbell, and including Georgia Tech researchers David A. Bader (School of Computational Science and Engineering), Mark Richards (School of Electrical and Computer Engineering), and Jeffrey Vetter (Oak Ridge National Laboratory and CSE, Georgia Tech) will focus UHPC architecture designs on achieving scalable capability on defense application requirements. CHASM will create high performance computing “challenge problems” that represent U.S. Department of Defense mission needs in the last half of the 2010s and will provide domain expertise on those problems and missions to the architecture teams. The CHASM team includes nationally recognized experts in each of the five challenge problem domains, which are:

• Streaming Sensors

• Dynamic Graph Analysis

• Decision/Search

• Lagrangian Shock Hydrodynamics

• Molecular Dynamics

These problems span a variety of application domains, computing styles, and performance requirements. CHASM will define scalable computing problems, tasks, and missions associated with each problem domain, create a written specification and a reference implementation, analyze the computing requirements, extract secondary metrics, and establish benchmarks.

ECHELONGeorgia Tech is a member of ECHELON (Extreme-scale Compute Hierarchies with Efficient Locality-Optimized Nodes), led by NVIDIA and including Cray, Lawrence Berkeley National Laboratory, Oak Ridge National Laboratory, and several other U.S. universities. The research team plans to develop new software and hardware technology to dramatically increase computing performance, programmability, and reliability. Design goals for ECHELON aim to support a wide range of workloads from massive data-intensive computing to compute-intensive simulations. The team proposes innovative general-purpose mechanisms that improve the energy efficiency, performance, resilience, and ease of programming, for challenging codes that span a wide variety of application areas and workflows. The ECHELON team’s Applications group, led by Georgia Tech’s David Bader, will investigate the UHPC challenge problems and use them to influence the Echelon system design.

UHPC PRO J EC TS

DARPA

Clockwise from top left: Dan Campbell, Mark Richards, Jeffrey Vetter, and David Bader

8 Georgia Tech [Institute for Data and High Performance Computing]

Research Areas• High Performance Computing

• Data Analytics, Machine Learning, and Visualization

• Modeling and Simulation

• Computational Mathematics

• Computational Science

• Computational Engineering

Research PartnersIDH faculty members collaborate with many companies, federal agencies, and universities on data intensive and high performance computing research. Select partners include Air Force Research Laboratory, Cray, Dell, U.S. Department of Defense, U.S. Department of Energy, Emory University, Hewlett-Packard, U.S. Department of Homeland Security, IBM, Intel, LexisNexis, Microsoft, National Institutes of Health, National Science Foundation, Northrop Grumman, NVIDIA, Oak Ridge National Laboratory, Pacific Northwest National Laboratory, Sandia National Laboratories, Sony, Sun Microsystems/Oracle, and Toshiba.

NextGen Codes InitiativeThe Georgia Tech NextGen Codes Initiative is stimulating the creation of new data and high performance computing codes. The initiative provides seed funding to grow and enhance multidisciplinary teams of researchers who will create the next generation of computational codes that will execute on emerging HPC platforms.

Four projects are currently funded in this program:

INSTITUTE FOR

DATA AND HIGH PERFORMANCE COMPUTING[ ]

• “PSI4: A Next Generation Computational Quantum Chemistry Package,” David Sherrill and Edmond Chow The goal of the PSI4 project is to develop efficient, parallel algorithms incorporating the very latest density-fitting, Cholesky decomposition, and localized-electron approximations. These algorithms will be integrated into a feature-rich quantum chemistry package that has been redesigned from the ground up as an object-oriented, massively parallel code.

• “Toward Exascale Pseudo-Spectral Codes for Turbulence Simulations on General-Purpose Graphics Processing Units,” Richard Vuduc and P. K. Yeung This project is developing new GPU-based codes for three-dimensional Fourier pseudo-spectral algorithms suitable for partial differential equations describing important phenomena in science and engineering. In particular, it is being applied to direct numerical simulation of turbulent fluid flow at the largest problem sizes possible. The project goal is to develop, test, and refine a new highly scalable GPU code for 3-D FFTs, at problem sizes 40963 and beyond.

• “MLPACK: Scalable Machine Learning Software for Large-Scale Science,” Alexander Gray, Alexander Shapiro, Haesun Park, Jeffrey Vetter, Richard Vuduc, John McDonald, Gordon Richards, Yu (Cathy) Jiao This project is building and disseminating a scalable, comprehensive, and usable machine learning library. Modeled on that of linear algebra’s LAPACK, MLPACK will be a comprehensive suite of state-of-the-art algorithms for applied mathematics and data analytics.

• “STING: A Software Framework for Dynamic Massive Graph Analysis,” Jason Riedy, David Bader, Dan Campbell This project is developing the first freely available, open-source framework for analysis of massive, dynamic Spatio-Temporal Interaction Networks and Graphs (STING). STING will provide not only a general framework but also specific analytical kernels to bootstrap rapid deployment and use.

About the School of Computational Science & EngineeringGeorgia Tech is devoted to the advancement and promotion of Computational Science & Engineering (CSE). Our approach differs from that of other universities. We view computational science and engineering as a discipline in its own right with its own distinct body of knowledge. We believe it should be represented on college campuses like other disciplines, such as mathematics or computer science: as a formal academic unit in order to create a natural home for faculty and students with expertise and interests in this field. Originally formed as a division in 2005, this philosophy led to the creation of the School of Computational Science and Engineering as a formal academic unit in 2010. We believe the CSE discipline fundamentally derives much of its richness and potential for impact from collaboration with other disciplines. Thus, the school has a strong emphasis on interdisciplinary research and education. CSE research spans many computational areas. For example, research in high performance computing develops new ways to exploit the world’s most powerful supercomputers. Research in massive scale data and visual analytics and machine learning explores ways to extract useful information from the unprecedented volumes of data now appearing on the Internet and in many fields of science, engineering, and medicine. Modeling and simulation research explores new methods to exploit parallel and distributed computing platforms in order to solve challenging problems in areas such as medicine and transportation. Algorithm research builds a solid foundation spanning both continuous and discrete models. Our research includes interdepartmental collaborations and interactions that crisscross the Georgia Tech campus—and extend around the world.

Degree ProgramsGeorgia Tech’s interdisciplinary degree programs develop a new type of scholar who is well versed in synthesizing principles from mathematics, science, engineering, and computing to create innovative computational models and apply them to solve important real-world problems. Computational Science and Engineering graduate degrees are jointly offered by the Colleges of Computing, Engineering, and Sciences at Georgia Tech.

• MS in Computational Science & Engineering (including a distance learning option)

• PhD in Computational Science & Engineering

What is Computational Science & Engineering?Computational Science and Engineering (CSE) is a discipline devoted to the systematic study, creation, and application of computer-based models to understand and analyze natural and engineered systems. It is inherently interdisciplinary, with close ties to other disciplines such as computer science, mathematics, science, and engineering. Subfields of CSE include high performance computing, data analytics and machine learning, visualization, modeling and simulation, and numerical and discrete algorithms. Computational modeling and data analytics are routinely used in virtually all fields of science and engineering to analyze systems as large as the universe and as small as the tiniest molecules. They are essential to solving the most important and challenging problems facing the world today, such as the diagnosis and prognosis of disease, the creation of sustainable cities, and the development of new sources of clean, inexpensive energy. In short, computational science and engineering has become indispensible in modern science and engineering.

idh.gatech.edu[ ]Institute for Data and High Performance Computing

Georgia Institute of Technology266 Ferst Drive

Klaus Advanced Computing BuildingAtlanta, GA 30332-0280

Phone: 404.385.4785Fax: 404.385.7337

Copyright 2010 • Georgia Institute of Technology • Communications & Marketing • B11C1004 • An equal education and employment opportunity institution