Putting this computing power into a small and energy-efficient package, and making it reliable and easier to program, are among the goals of the new DARPA Ubiquitous High Performance Computing (UHPC) initiative. Georgia Tech researchers from three different units are supporting key components of this $100 million challenge, which will require development of revolutionary approaches not bound by existing computing paradigms.
If UHPC meets its ambitious eight-year goals, the new approaches and technologies it develops could redefine the way that computing systems are envisioned, designed and used.
"The opportunity we have is to go far beyond the current product roadmaps," said David Bader, a professor in Georgia Tech's School of Computational Science and Engineering. "We really have the opportunity to change the industry and to design our applications with new computing architectures. For the first time in the history of computing, we will be able to work with a clean slate."
To attain the program's ambitious goals, DARPA funded four groups -- led by NVIDIA Corp., Intel Corp., the Massachusetts Institute of Technology and Sandia National Laboratories -- to develop UHPC prototypes. A fifth group, led by the Georgia Tech Research Institute (GTRI), will develop applications, benchmarking and metrics that will be used to drive UHPC system design considerations and support performance analysis of the developing system designs.
"Our team is developing a set of five difficult problems of a size and scope that the machines they are talking about should be able to accomplish," said Dan Campbell, a GTRI principal research engineer who is co-principal investigator of the benchmarking initiative. "Our challenge is picking the right problems and specifying them at the right level of abstraction to allow innovation and properly represent what the DoD will need in 2018."
The five problems highlight the unique computing needs of the U.S. military:
• Analysis of the vast streams of data originating with widespread sensor systems, unmanned aerial vehicles and new generations of radar systems. The data will be analyzed for nuggets of useful information in ways that are not possible today.
• A dynamic graph challenge, in which many entities interact to create a problem of "connecting the dots." That could mean analyzing relationships in social media to find possible adversaries, or understanding network traffic for cyber-security challenges.
• The decision tree, comparable to a chess game in which many possible interconnected options, each with complex implications, must be analyzed quickly. This could help field commanders or corporate CEOs make better decisions.
• Materials shock and hydrodynamics issues, challenges important to improving future generations of materials.
• Molecular dynamics simulations, which use high-performance computers to understand interactions between very large systems, such as protein folding.
"We need to be able to take in a lot more data and understand it a lot more thoroughly than we can now," said Mark Richards, a principal research engineer in the Georgia Tech School of Electrical and Computer Engineering and co-principal investigator of the benchmarking team. "That might allow us to find adversaries we can't find now because we're unable to tease that information out of the data flow."
While the benefits of making such computing power widely available are obvious, how these machines will be designed, built and reliably operated is not.
"Meeting these very ambitious program goals will pose significant technical challenges," said Bader, who leads application development on the NVIDIA team and is part of the benchmarking group. "The technology roadmaps in such areas as interconnection networks, microprocessor design and technology fabrication will be pushed to their limits."
Meeting power limitations of just 57 kilowatts per rack -- the amount of electricity produced by a portable military generator -- may be the toughest among them. The fastest computer currently in operation requires seven megawatts of power.
"Reducing the power consumption means less energy per computation," noted Richards. "But as we lower the device voltage, we get closer to the physical noise. That will allow more errors due to the physics of the devices, and all kinds of things will have to be done to address that."
And the entire machine will have to fit into a 24-inch wide, 78-inch high and 40-inch deep cabinet.
But the physical implementation of the machines is just one part of the challenge, Bader noted. How people will work with them poses a perhaps more difficult challenge because it will require thinking about computers in a new way.
"Over the past 20 or 30 years, we've taken a single computing design and kept tweaking it through advances like miniaturizing parts," he said. "But we really haven't changed the global nature of how the machine works. To meet DARPA's power efficiency goals, we really will need to change the way we program the machine."
That also affects the humans who interact with these highly-parallel machines, which could have as many as a half-million separate threads operating at the same time. DARPA's initial goal is to build machines capable of petaflop speed -- a trillion operations per second -- which could lead into the next generation of exascale computers a thousand times more capable.
"We will need to find new ways of thinking about computers that will make it feasible for humans to comprehend what is going on inside," Campbell said. "It's a huge programming challenge."
To encourage collaboration in solving these complex problems, DARPA has embraced the idea of open innovation. It expects the organizations to work together on common critical topics, creating a collaborative environment to address the system challenges. New technology generated by the program -- believed to be today's largest DoD computing research initiative -- is likely to move quickly into industry.
"There is certainly an expectation among the companies that what they are doing in this project is going to change how we do mainstream computing," Bader said. "The technology transfer implications are certainly obvious."
]]>College of Computing
404-312-6620
]]>The contract is part of DARPA’s XDATA program, a four-year research effort to develop new computational techniques and open-source software tools for processing and analyzing data, motivated by defense needs. Georgia Tech has been selected by DARPA to perform research in the area of scalable analytics and data-processing technology.
The Georgia Tech team will focus on producing novel machine-learning approaches capable of analyzing very large-scale data. In addition, team members will pursue development of distributed computing methods that can process data-analytics algorithms very rapidly by simultaneously utilizing a variety of systems, including supercomputers, parallel-processing environments and networked distributed computing systems.
"This award allows us to build on the foundations we've already established in large-scale data analytics and visualization," said Richard Fujimoto, Regents' Professor and chair of Georgia Tech’s School of Computational Science and Engineering (CSE), and leader of the Georgia Tech team. "The algorithms, tools and other technologies that we develop will all be open source, to allow them to be customized to address new problems arising in defense and other applications."
Under the open-source paradigm, collaborating developers create and maintain software and associated tools. Program source code is made widely available and can be improved by a community of developers and modified to address changing needs.
The XDATA award is part of a $200 million multi-agency federal initiative for big-data research and development announced in March. The initiative is aimed at improving the ability to extract knowledge and insights from the nation's fast-growing volumes of digital data. Numerous big-data-related research endeavors are underway at Georgia Tech, and the institute recently established the Center for High-Performance Computing and the Center for Data Analytics and Machine Learning.
The Georgia Tech XDATA effort will build upon foundational methods and software developed under the Foundations of Data and Visual Analytics (FODAVA) research initiative, a 17-university program led by Georgia Tech and funded by the National Science Foundation and the Department of Homeland Security. The FODAVA effort has produced the Visual Information Retrieval and Recommendation System (VIZIRR) and a research test bed.
"The FODAVA document retrieval and recommendation system uses automated algorithms to give users a range of subject-search choices and information visualization capabilities in an integrated way, so that users can interact with the data throughout the problem-solving process to produce more meaningful solutions," said Haesun Park, a School of Computational Science and Engineering professor and FODAVA director. "For XDATA, we will enhance these visualization and interaction capabilities and develop distributed algorithms that allow users to solve problems faster and on a larger scale than ever before."
Also participating from the School of Computational Science and Engineering is Alex Gray, an associate professor who has developed open-source software tools to make machine-learning algorithms scalable to large datasets. Other faculty members involved in the XDATA work include professor Hongyuan Zha and associate professor Guy Lebanon.
Investigators from the Georgia Tech Research Institute (GTRI) will also contribute to the XDATA initiative. Senior research scientists Barry Drake and Richard Boyd will tackle the computational demands of processing the machine-learning algorithms developed by the School of Computational Science and Engineering team.
GTRI's task involves enabling these algorithms to run on a networked distributed computing system. By configuring the software so that it operates on multiple processors simultaneously, the researchers believe they can ensure that the algorithms solve problems very rapidly – a requirement of the DARPA award.
"Scaling up machine-learning algorithms to big-data requirements is a relatively new area of research, and there will be both hardware and software issues to address here," said Drake, a specialist in parallel algorithms in numerous application domains. "In enabling these complex codes to analyze large data sets rapidly, we expect to be breaking new ground."
Boyd will support XDATA's hardware requirements with expertise on low-cost graphics processing units (GPUs), which offer performance levels reached only by supercomputers until recently. Clusters of linked GPUs could help provide the processing power needed to satisfy XDATA requirements.
"The XDATA vision involves providing an entirely new set of open-source data-processing tools for both military and other requirements," Boyd said. "We have to be prepared to deal with not only widely distributed computing, but also with heterogeneous data that could be structured or unstructured. Diverse hardware approaches including GPUs are likely to be part of the system."
For more information on DARPA and the XDATA program, visit www.darpa.mil.
Research News & Publications Office
Georgia Institute of Technology
177 North Avenue
Atlanta, Georgia 30332-0181
Media Relations Contacts: John Toon (404-894-6986)(jtoon@gatech.edu), Joshua Preston (404-385-3845)(jpreston@cc.gatech.edu) or Lance Wallace (404-407-7280)(lance.wallace@gtri.gatech.edu).
Writer: Rick Robinson
Research News & Publications Office
(404) 894-6986
]]>Research in computational science and engineering at Georgia Tech spans many areas ranging from the development of new computational methods that may be applied to one or more fields in science and engineering to novel computational approaches specific to a particular domain such as biology or aerospace engineering.
Because Georgia Tech views computation as the driver of future advances in science and engineering, the School of Computational Science and Engineering was created to be a truly interdisciplinary unit that crosses the conventional academic boundaries found between research disciplines. Faculty from all walks of computing, sciences, and engineering collaborate within six core areas: High-Performance Computing; Data Analytics, Machine Learning and Visualization; Modeling and Simulation; Computational Mathematics; Computational Science; and Computational Engineering.
The HPC500 is comprised of a representative cross-section of academic, government, and commercial organizations across all budgets, applications, and geographic areas, including users in both High Performance Technical Computing (HPTC) and High Performance Business Computing (HPBC). The charter members are listed at the HPC500 Website (http://www.hpc500.com/member-directory/).
Of the first fifty members:
For more information about High Performance Computing at Georgia Tech, please contact David A. Bader, professor in the School of Computational Science and Engineering and executive director of High Performance Computing, at 404-894-5756.
]]>678.231.0787
]]>Kernel summations are a ubiquitous key computational bottleneck in many data analysis methods. The paper proposes a hybrid MPI/OpenMP kernel summation framework for scaling many popular data analysis methods. Advantages to the approach include utilizing the platform-independent C++ code base that utilizes standard protocols such as MPI and OpenMP; using the template code structure that uses any multidimensional binary trees and any approximation schemes that may be suitable for high-dimensional problems; and having extendibility to a large class of problems that require fast evaluations of kernel sums.
“Researchers have previously parallelized kernel summations in the context of simulations,” says Dongryeol Lee, a Ph.D. candidate in Computer Science. “But this paper is the first serious effort in parallelizing kernel summations in the context of data mining with potentially high-profile scientific applications.”
In data mining, kernel summations appear in popular so-called kernel methods which can model complex, nonlinear structures in data. The richer expressiveness of the methods comes with the drawback of requiring many data points and hence more computational power for crunching collected data, according to Lee. The collected data in some cases must be stored on multiple machines.
From the data mining community, Lee says this work is the first to utilize algorithmic techniques in both high performance computing, computer science, computational physics, computational geometry, and approximation theory in a general framework.
Kernel summations drive algorithms in application areas such as finance, astronomy, and medical science.
Lee notes some examples: “Fraudulent financial transactions can be detected more quickly using fast kernel summations. Astronomy uses the algorithms to predict redshift of many galaxies and stars, which can shed light onto the ultimate fate of the universe. Medicine uses fast kernel summation algorithms in automated early detection of cancer that can save human lives."
]]>678-231-0787
]]>
The NMSC focused on establishing, for the first time, a national agenda for maintaining the growth of modeling and simulation technology and its incorporation into all areas of the national economy, welfare, and security.
Richard Fujimoto, chair of the School of Computational Science and Engineering and interim director of the Institute for Data and High Performance Computing at the Georgia Institute of Technology, is a member of the interim Board of Directors and interim chair of the Education and Professional Development Standing Committee, one of five standing committees of the National Modeling and Simulation Coalition. He chaired the inaugural meeting for this committee at the Feb. 6 event.
“A critical issue that the committee is starting to address is how to increase the number of people trained and educated in modeling and simulation in order to meet the high workforce demand,” Fujimoto said. “We are developing a national agenda that will focus in part on this issue.”
The committee’s work will span the entire education pipeline, including K-12, higher education, technical education and continuing education.
Within the School of Computational Science and Engineering at Georgia Tech, the approach is to develop education programs in computing for students – who are studying engineering, the sciences and other areas - that build up their computing capability, said Fujimoto.
“We really must be focused on developing workforce needs,” the CSE chair said.
At the event in Washington, D.C., Aneesh Chopra, chief technology officer for the United States, gave the keynote address after a ceremonial ribbon cutting for the NMSC.
The event featured several hundred key leaders from industry, government, and academia. Also attending from Georgia Tech was Margaret Loper, a chief scientist in the Georgia Tech Research Institute.
Participants set the stage for the initial organizational and committee sessions and to define a detailed action plan in four key areas: Education and Professional Development, Technology Research and Development, Industrial Development, and Business Practice. The outcome of the inaugural meeting included a national action plan to broaden the use of modeling and simulation across these four key areas. This action plan will provide a road map for the modeling and simulation community and in expanding the depth and breadth of the technology and industry.
]]>404-385-3845
]]>]]>
]]>
Source: The Hill's Congress Blog
Source: HPCWire
]]>Georgia Tech will develop technologies to accelerate the design and manufacture of high performance materials for applications - ranging from fuel-efficient vehicles to emerging technologies such as 3D printing – through the new graduate training program.
“The goal of this program is to employ advances in ‘big data’ and information technology to significantly reduce the timelines now required for new materials to be created and incorporated into commercial products,” said School of Computational Science and Engineering Chair and Regents’ Professor Richard Fujimoto, the principal investigator for the grant.
“The program will be transformational in bringing ‘big data’ researchers together with materials scientists, engineers, and mathematicians to quantify the microstructures that comprise materials and develop new algorithms and software for their design,” said Fujimoto, who leads Georgia Tech’s Institute for Data and High Performance Computing (IDH).
The new Georgia Tech program includes a focus on entrepreneurship to enable graduate trainees to transform technical innovations into commercial products and services. Called FLAMEL - From Learning, Analytics, and Materials to Entrepreneurship and Leadership, the program references 15th-century alchemist Nicolas Flamel, known as the creator of the philosopher’s stone, a substance purported to transform materials into gold.
“We hope to boost the United States’ competitiveness in an evolving global marketplace by enabling our engineers and scientists to explore the potential commercialization of their ideas and inventions,” says Terry Blum, a co-investigator of the program and the founding director of Georgia Tech’s Institute for Leadership and Entrepreneurship.
FLAMEL is aligned with the federal Materials Genome Initiative, focused on cutting the development time for advanced materials in half and reducing their costs. FLAMEL will leverage Georgia Tech’s recent investment in MatIN, a cyberinfrastructure platform designed to enable rapid interdisciplinary collaboration in materials development and manufacture. MatIN is being currently developed as a joint collaboration between Georgia Tech’s Institute for Materials (IMat), Institute for Data and High Performance Computing (IDH), Office of Information Technology (OIT), and Georgia Tech Research Institute (GTRI).
Central to the program will be an emerging area known as materials informatics, which aims to develop new approaches to materials design and manufacturing using data analytics combined with modeling and simulation, according to Surya Kalidindi, program co-deputy director, and member of the IMat Innovation support team.
Other members of the Georgia Tech team include co-investigator Wendy Newstetter, a learning scientist and leader in the development and use of problem-based learning methods; and co-deputy director Hongyuan Zha, an expert in “big data” algorithms and software.
While the emphasis will be on doctoral students, the program is designed to create a pipeline for workforce development that includes a broadened participation of women and minority students. The five-year program will provide funding for 24 doctoral trainees but is expected to create educational opportunities that will impact hundreds of Georgia Tech students in the years ahead.
For application materials or more information about the FLAMEL program, visit flamel.gatech.edu or contact program coordinator Holly Rush at holly@cc.gatech.edu.
The computationally focused training grant is funded through NSF’s Graduate Education and Research Traineeship (IGERT) program, award number DGE-1258425. Any opinions or conclusions expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.
]]>“For example, we get together with our friends to figure out which houses to go to for optimal candy rewards — and optimization is one component of graph theory,” said Bader, professor in the College of Computing and executive director for High Performance Computing. “And when you get home, you sort your candy — yet another concept. We should consider October graph theory month.”
Welcome to Bader’s Computational Science and Engineering Algorithms class — and his approach to teaching.
“When I prepare for class, I think about the story I’m going to tell and try to incorporate current events and real-world cases into the curriculum,” he said.
But these skills didn’t develop overnight.
From Techie Toddler to Full Prof
It began in 1972, when Bader started playing around on a mainframe computer at the college where his dad worked. He was 3.
“Then, when I was in elementary school in the late 1970s, I began holding computer retraining classes for the parents of my friends and neighbors who were being laid off from a nearby steel plant and were looking for a new profession,” he added. “Things just seemed to evolve from there.”
After spending time at the University of Maryland as a National Science Foundation postdoctoral research associate, Bader began working his way through the professorial ranks at the University of New Mexico in the late 1990s. He was recruited to Georgia Tech in 2005 to launch the School of Computational Science and Engineering.
Bader’s research aims to solve real-world problems by using computational and data-intensive solutions.
For example, Bader might use massive data sets to help him figure out how to keep people safer during severe weather or how to design more personalized medications.
“I love that my job is fun,” he said. “I can’t believe I get paid to collaborate with the most amazing students and colleagues, and solve grand challenge problems that have a lasting impact on society.”
The Man Behind the Monitor
When Bader isn’t conducting research or teaching, you might find him enjoying one of his favorite spots on campus, the Klaus Advanced Computing Building.
“It’s just a beautiful facility that inspires me to make contributions to computing and to think about the impact of my research on society,” he added.
Or you might catch this vegan enjoying the various dining spots around campus to try their takes on international cuisines ranging from African to Asian.
And one last fun fact about this computing wiz — he also has a strong interest in genomics (simply put, the study of genes and their functions).
“Many people might not be aware that I have a twin sister — who happened to earn her doctorate in genetics,” he said. “My full mitochondrial genome is even available in GenBank, a database that contains publicly available DNA sequences.”
]]>That’s about to change.
Researchers at the Georgia Institute of Technology and University of Southern California recently received a nearly $2 million federal grant to develop tools to assist all developers in using hardware accelerators productively and effectively.
The goal of the three-year grant from the National Science Foundation is to bring capability to a general audience, including those using tablets, smartphones and other ubiquitous devices, said David Bader, the lead principal investigator.
Imagine an elderly woman using her smartphone to understand her health informatics or farmers in developing nations using hardware accelerators to understand weather patterns so they know when to plant crops.
“We want to take science that used to be only available to elite scientists and bring that to everyone around the planet,” said Bader, a professor in Georgia Tech’s School of Computational Science and Engineering and executive director for High Performance Computing. “We are bringing supercomputing to the masses.”
Bader will collaborate with Viktor Prasanna, another principal investigator who is a professor of electrical engineering and professor of computer science at the University of Southern California. Prasanna is executive director of the USC-Infosys Center for Advanced Software Technologies.
Rich Vuduc, an associate professor in the School of Computational Science and Engineering at Georgia Tech, and Jason Riedy, a research scientist, are co-principal investigators.
The grant focuses on two key areas. The first is to develop XScala, a software framework for developing efficient accelerator kernels. Data-intensive kernels include large dictionary string matching, dynamic programming and graph theory.
Over the course of three years researchers will focus on different types of optimizations. They will emphasize power efficiency – an area with little current research – and hope to make discoveries that will allow for the development of new types of algorithms for accelerators, Bader said.
Early research efforts will focus on key areas such as security and social network analysis. The impact of this early research will be felt in other areas because the algorithms from these domains can be applied to other science and technology segments that use graph analytics.
The second focus calls for a public software repository and forum, called XBazaar. Similar to the iPhone App Store, XBazaar will serve as a one-stop shop for high-performance algorithms and software for multi-core and many-core processors.
This open source will create a marketplace to develop and share best practices for using accelerators. People will be able to modify code or use ideas from the researchers’ code to develop their own high-performance applications that are cost-effective and power-efficient, Bader said.
XBazaar can be used in commercial applications, allowing it to spur economic development and benefit industry.
“Our goal is to build something larger than what we can do ourselves,” Bader said.
]]>This annual award recognizes one individual based on their overall contributions to the field of modeling and simulation, including technical innovations, publications, leadership, teaching, mentoring, and overall service to the community.
Fujimoto is a Regents' Professor in the College of Computing at the Georgia Institute of Technology. He is also the chair of the School of Computational Science and Engineering and the interim director of the Institute for Data & High Performance Computing (IDH).
The ACM Special Interest Group on Simulation and Modeling seeks the advancement of the state-of-the-art in simulation and modeling. Simulation and modeling cut across a broad range of interests. SIGSIM joins each year with a variety of other organizations in co-sponsoring the Winter Simulation Conference, the premier conference in the field, and the Parallel and Distributed Simulation Conference (PADS).
]]>The 2014 list is a compilation of the 16 best and brightest minds from academia, science, and technology whose contributions in high performance computing (HPC) have the potential to profoundly impact the world this year and beyond. Finalists are selected following an extensive review process by the HPCwire editorial and executive staff along with guidance from industry analysts and luminaries across the HPC community.
Although HPCwire has recognized Bader as a “person to watch” before, he continues to garner attention and recognition for his work in big data at the intersection where high performance computing meets real-world applications.
Among his many achievements, Bader led an effort that resulted in a nearly $2 million dollar grant for researchers at Georgia Tech and the University of Southern California to bring supercomputing capabilities into the grasp of tablets, smart phones, and other devices.
Read more about Bader and his thoughts on the trends in HPC for 2014.
]]>Google has been working on artificial intelligence almost since its founding 16 years ago. Its efforts focused on improving search results, translation, and filtering spam. Google Now, a digital personal assistant within Google's mobile search app, is among the most ambitious of its efforts to date. It guesses what information users want based on their past search history and location, and then gives it to them.
"Google Now can do simple things like give you directions to work or figure out that you have a flight," Bader said. "But to really predict what you're really going to do or to understand your
]]>The new system is called LatentGesture and was used during a Georgia Tech lab study using Android devices. The system was nearly 98 percent accurate on a smartphone and 97 percent correct on tablets. The research team will present the findings for the first time at the end of April.
“The system learns a person’s ‘touch signature,’ then constantly compares it to how the current user is interacting with the device,” said Polo Chau, a Georgia Tech College of Computing assistant professor who led the study.
To test the system, Chau and his team set up an electronic form with a list of tasks for 20 participants. They were asked to tap buttons, check boxes and swipe slider bars on a phone and tablet to fill out the form. The system tracked their tendencies and created a profile for each person.
After profiles were stored, the researchers designated one person’s signature as the “owner” of the device and repeated the tests. LatentGesture successfully matched the owner and flagged everyone else as unauthorized users.
“Just like your fingerprint, everyone is unique when they use a touchscreen,” said Chau. “Some people slide the bar with one quick swipe. Others gradually move it across the screen. Everyone taps the screen with different pressures while checking boxes.”
The research team also programmed the system to store five touch signatures on the same device – one “owner” and four authorized users. When someone other than the owner used the tablet, the system identified each with 98 percent accuracy.
“This feature could be used when a child uses her dad’s tablet,” said College of Computing sophomore Premkumar Saravanan. “The system would recognize her touch signature and allow her to use the device. But if she tried to buy an app, the system could prevent it.”
The researchers say LatentGesture’s biggest advantage is that the system is constantly running in the background. The user doesn’t have to do anything different for added security and authentication.
“It’s pretty easy for someone to look over your shoulder while you’re unlocking your phone and see your password,” said Samuel Clarke, another College of Computing student on the research team. “This system ensures security even if someone takes your phone or tablet and starts using it.”
Chau is co-advising the project with Hongyuan Zha, a professor in the School of Computational Science and Engineering. The study will be presented in Toronto at ACM Chinese CHI 2014 from April 26 to 27.
This research was partially supported by the National Science Foundation (NSF) under grants IIS-1049694 and IIS-1116886. Any conclusions expressed are those of the principal investigator and may not necessarily represent the official views of the NSF.
]]>Modern computing systems, to meet scientific demands of the future, cannot rely solely on the expanded computing power of hardware but also need algorithms and software that can efficiently use massive amounts of parallelism. Intel is creating Intel Parallel Computing Centers (IPCCs) at leading institutions in HPC research to promote the modernization of essential application codes to increase their parallelism and scalability.
The IPCC at Georgia Tech will develop new parallel algorithms and software for quantum chemistry and biomolecular simulation. Research will target large-scale computer systems using Intel Xeon Processors and Intel Xeon Phi coprocessors. The center will also develop new curricular materials to equip future computer scientists with the skills to fully realize the capabilities of parallel computing resources for scientific applications.
“We're thrilled for Georgia Tech to be named as the next IPCC,” said Edmond Chow, associate professor and principal investigator of the new IPCC. “In the College of Computing we tackle large-scale problems with a holistic approach, combining high-performance computing with deep knowledge in scientific applications as well as novel algorithms based in applied mathematics. This collaboration with Intel gives us access to their significant expertise in hardware systems and their optimization, and allows us to innovate in a multi-disciplinary way that we could not do before.”
“Georgia Tech is doing top-notch research in high performance computing,” said Pradeep Dubey, Intel fellow and director of the Intel Parallel Computing Lab. “Intel has had a long history of productive collaboration with the School of Computational Science and Engineering. We expect research at this IPCC to significantly advance the capabilities of supercomputers as they approach the frontiers of Exascale computing.”
]]>News & Media Relations Manager
404.894.7253
]]>In a universe with tens of billions of galaxies, how would we see it? What would such a beacon look like? And how would we distinguish it from other bright, monumental intergalactic events, such as supernovas?
"Black holes by themselves do not emit light," said Tamara Bogdanovic, an assistant professor of physics at the Georgia Institute of Technology. "Our best chance to discover them in distant galaxies is if they interact with the stars and gas that are around them."
In recent decades, with improved telescopes and observational techniques designed to repeatedly survey the vast numbers of galaxies in the sky, scientists noticed that some galaxies that previously looked inactive would suddenly light up at their very center.
"This flare of light was found to have a characteristic behavior as a function of time. It starts very bright and its luminosity then decreases in time in a particular way," she explained. "Astronomers have identified those as galaxies where a central black hole just disrupted and 'ate' a star. It's like a black hole putting up a sign that says 'Here I am.'"
Using a mix of theoretical and computer-based approaches, Bogdanovic tries to predict the dynamics of events such as the black-hole-devouring-star scenario described above, also known as a "tidal disruption." Such events would have a distinct signature to someone analyzing data from a ground-based or space-based observatory.
Using National Science Foundation-funded supercomputers at the Texas Advanced Computing Center (Stampede) and the National Institute for Computational Sciences (Kraken), Bogdanovic and her collaborators recently simulated the dynamics of these super powerful forces and charted their behavior using numerical models.
Tidal disruptions are relatively rare cosmic occurrences. Astrophysicists have calculated that a Milky Way-like galaxy stages the disruption of a star only once in about 10,000 years. The luminous flare of light, on the other hand, can fade away in only a few years. Because it is such a challenge to pinpoint tidal disruptions in the sky, astronomical surveys that monitor vast numbers of galaxies simultaneously are crucial.
Huge difference
So far, only a few dozen of these characteristic flare signatures have been observed and deemed "candidates" for tidal disruptions. But with data from PanSTARRS, Galex, the Palomar Transient Factory and other upcoming astronomical surveys becoming available to scientists, Bogdanovic believes this situation will change dramatically.
"As opposed to a few dozen that have been found over the past 10 years, now imagine hundreds per year--that's a huge difference!" she said. "It means that we will be able to build a varied sample of stars of different types being disrupted by supermassive black holes."
With hundreds of such events to explore, astrophysicists' understanding of black holes and the stars around them would advance by leaps and bounds, helping determine some key aspects of galactic physics.
"A diversity in the type of disrupted stars tells us something about the makeup of the star clusters in the centers of galaxies," Bodganovic said. "It may give us an idea about how many main sequence stars, how many red giants, or white dwarf stars are there on average."
Tidal disruptions also tell us something about the population and properties of supermassive black holes that are doing the disrupting.
"We use these observations as a window of opportunity to learn important things about the black holes and their host galaxies," she continued. "Once the tidal disruption flare dims below some threshold luminosity that can be seen in observations, the window closes for that particular galaxy."
Role of supercomputer
In a recent paper submitted to the Astrophysical Journal, Bogdanovic, working with Roseanne Cheng (Center for Relativistic Astrophysics at Georgia Tech) and Pau Amaro-Seoane (Albert Einstein Institute in Potsdam, Germany), considered the tidal disruption of a red giant star by a supermassive black hole using computer modeling.
The paper comes on the heels of the discovery of a tidal disruption event in which a black hole disrupted a helium-rich stellar core, thought to be a remnant of a red giant star, named PS1-10jh, 2.7 billion light years from Earth.
The sequence of events they described aims to explain some unusual aspects of the observational signatures associated with this event, such as the absence of the hydrogen emission lines from the spectrum of PS1-10jh.
As a follow-up to this theoretical study, the team has been running simulations on Kraken and Stampede, as well as the Georgia Tech's high performance computing clusters. The simulations reconstruct the chain of events by which a stellar core, similar to the remnant of a tidally disrupted red giant star, might evolve under the gravitational tides of a massive black hole.
"Calculating the messy interplay between hydrodynamics and gravity is feasible on a human timescale only with a supercomputer," Cheng said. "Because we have control over this virtual experiment and can repeat it, fast forward, or rewind as needed, we can examine the tidal disruption process from many perspectives. This in turn allows us to determine and quantify the most important physical processes at play."
The research shows how supercomputer simulations complement and constrain theory and observation.
"There are many situations in astrophysics where we cannot get insight into a sequence of events that played out without simulations. We cannot stand next to the black hole and look at how it accretes gas. So we use simulations to learn about these distant and extreme environments," Bogdanovic said.
One of Bogdanovic's goals is to use the knowledge gained from simulations to decode the signatures of observed tidal disruption events.
"The most recent data on tidal disruption events is already outpacing theoretical understanding and calling for the development of a new generation of models," she explained. "The new, better quality data indicates that there is a great diversity among the tidal disruption candidates. This is contrary to our perception, based on earlier epochs of observation, that they are a relatively uniform class of events. We have yet to understand what causes these differences in observational appearance, and computer simulations are guaranteed to be an important part of this journey."
-- Written by Aaron Dubrow of the National Science Foundation.
]]>
Research News
(404) 894-6986
]]>The Institute of Electrical and Electronics Engineers (IEEE) International Parallel & Distributed Processing Symposium (IPDPS), held from May 19 to 23 in Phoenix, is the flagship activity of the IEEE Computer Society’s Technical Committee on Parallel Processing (TCPP), representing a unique international gathering of computer scientists from around the world.
For IPDPS 2014, Georgia Tech participates in virtually every part of the technical program, where researchers from the College of Computing and the Institute for Data and High Performance Computing (IDH) have gathered to present their latest research findings in all aspects of parallel computation. They will take part in numerous paper presentations, workshops, and other parts of the 28th annual IPDPS program. David A. Bader serves as the 2014 IPDPS program chair, and Srinivas Aluru has been tapped as the program chair for the 2015 symposium.
This year, Georgia Tech researchers—CSE’s Associate Professor Edmond Chow and graduate students Xing Liu and Aftab Patel, and CS’s Associate Professor Santosh Pande and graduate student Kaushik Ravichandran—took home two of the four best paper awards given at IPDPS 2014.
Georgia Tech’s participation at IPDPS 2014 includes:
Zhihui Du (visiting professor)
A New Scalable Parallel Algorithm for Fock Matrix Construction
Xing Liu (Georgia Institute of Technology, USA); Aftab Patel (Georgia Institute of Technology, USA); Edmond Chow (Georgia Institute of Technology, USA)
F2C2-STM: Flux-based Feedback-driven Concurrency Control for STMs
Kaushik Ravichandran (Georgia Institute of Technology, USA); Santosh Pande (Georgia Institute of Technology, USA)
Scibox: Online Sharing of Scientific Data via the Cloud
Jian Huang (Georgia Institute of Technology, USA); Xuechen Zhang (Georgia Institute of Technology, USA); Greg Eisenhauer (Georgia Institute of Technology, USA); Karsten Schwan (Georgia Tech, USA); Matthew Wolf (Georgia Institute of Technology, USA); Stephane Ethier (PPPL, USA); Scott Klasky (Oak Ridge National Laboratory, USA)
Parallel Mutual Information Based Construction of Whole-Genome Networks on the Intel(R) Xeon Phi(TM) Coprocessor
Sanchit Misra (Intel Corporation, India); Kiran Pamnany (Intel Corporation, India); Srinivas Aluru (Georgia Institute of Technology, USA)
Finding Motifs in Biological Sequences Using the Micron Automata Processor
Indranil Roy (Georgia Institute of Technology, USA); Srinivas Aluru (Georgia Institute of Technology, USA)
TBPoint: Reducing Simulation Time for Large Scale GPGPU Kernels
Jen-Cheng Huang (Georgia Institute of Technology, USA); Lifeng Nai (Georgia Institute of Technology, USA); Hyesoon Kim (Georgia Tech, USA); Hsien-Hsin Lee (Georgia Institute of Technology, USA)
Algorithmic Time, Energy, and Power on Candidate HPC Compute Building Blocks
Jee Choi (Georgia Institute of Technology, USA); Marat Dukhan (Georgia Institute of Technology, USA); Xing Liu (Georgia Institute of Technology, USA); Richard W. Vuduc (Georgia Institute of Technology, USA)
Large-scale Hydrodynamic Brownian Simulations on Multicore and Manycore Architectures
Xing Liu (Georgia Institute of Technology, USA); Edmond Chow (Georgia Institute of Technology, USA)
Interactive Program Debugging and Optimization for Directive-Based, Efficient GPU Computing
Seyong Lee (Oak Ridge National Laboratory, USA); Dong Li (Oak Ridge National Laboratory, USA); Jeffrey S. Vetter (Oak Ridge National Laboratory & Georgia Tech, USA)
Presented in Conjunction with IPDPS 2014.
Parallelization of the Trinity Pipeline for de Novo Transcriptome Assembly
Vipin Sachdeva (Georgia Institute of Technology, USA); Chang-Sik Kim, Kirk Jordan; Martyn D. Winn
Revisiting Edge and Node Parallelism for Dynamic GPU Graph Analytics
Adam McLaughlin (Georgia Institute of Technology, USA); David A. Bader (Georgia Institute of Technology, USA)
The students talked with farmers and volunteers about the crops, planting schedules, harvest requests, visitor demographics and other data crucial to the daily operation.
Urban agriculture, the students realized, is a complex undertaking. Their challenge is to create a streamlined data management system for the farm and move them away from pencil and paper. Ideally, this system will allow the farm to increase productivity and move toward financial sustainability.
The student team is one of five working with non-profits and government agencies as part of the Data Science for Social Good internship program, sponsored by Georgia Tech and Oracle.
Sixteen students from around the country are participating in a 10-week paid internship program showing non-profits and government agencies how they can use data to tackle social and societal problems.
The program allows students to solve real-world problems instead of relying on sample data sets, said Ellen Zegura, the program director.
It also educates local non-profits on the need for better data systems, said Zegura, a professor in the School of Computer Science in the College of Computing at Georgia Tech.
“We are connecting those who collect data with the people who know how to turn the data into something meaningful that can have a positive impact,” she said.
The projects deal with safety, criminal justice, transportation and sustainability. The student teams are collaborating with the Atlanta Police Department, the city’s Community Courts, Cycle Atlanta, Georgia Tech Office of Information Technology and Truly Living Well.
Georgia Tech is piloting the program this year and hopes to grow it next year. The Atlanta internship is modeled after a similar program the University of Chicago started last year.
Raj Bandyopadhyay, principal data scientist with Pindrop Security, heard about the Chicago program at a conference and led the charge to bring it to Atlanta.
“So often when people hear of big data, they don’t understand how it can be used to improve their lives,” he said. “We are showing future data scientists how they can use their skills to address social issues.”
More than 80 students applied for the internship. The selected students come from eight colleges including: Carnegie Mellon University, Southern Methodist University, Emory University and Georgia Tech.
Umashanthi Pavalanathan, a Ph.D. student in computer science at Georgia Tech, is working with 911 data collected by the city. They are looking at the response time between calls and how to best use the dispatchers.
“I’m so used to dealing with abstract concepts and situations that it’s exciting to work with real clients on real issues,” she said. “You get a good feeling knowing that what we’re doing can help save somebody.”
Students from all five projects will present their findings and recommendations during a public demonstration and reception scheduled to begin at 6:30 p.m. on July 17 at Atlanta Tech Village. More information about the event and internship program can be found here: http://dssg-atl.io
The internship ends July 18, although the data will continue to reap benefits.
]]>“The program is looking at how do we come to a new paradigm of computing where running time isn’t necessarily the constraint, but how much power and battery that we have available is really the new constraint,” says David Bader, executive director of high-performance computing at the School of Computational Science and Engineering.
If the project is successful, it could result in computers far smaller and orders of magnitude more efficient than today’s machines. It could also mean that the computer mounted tomorrow on an unmanned aircraft or ground vehicle, or even worn by a soldier would use less energy than a larger device, while still being as powerful.
Georgia Tech’s part in the DARPA-led PERFECT effort is called GRATEFUL, which stands for Graph Analysis Tackling power-Efficiency, Uncertainty and Locality. Headed by Bader and co-investigator Jason Riedy, GRATEFUL focuses on algorithms that would process vast stores of data and turn it into a graphical representation in the most energy-efficient way possible.
The ultimate goal is to get an algorithmic framework that delivers supercomputer capabilities on a small, power-restricted platform.
One approach to reducing power consumption is to reduce the level of data collection. For example, when looking for a needle in a haystack, you don’t necessarily need to inspect every piece of hay. “What we’re looking at is collecting the minimal data necessary to make accurate decisions,” Bader says.
For now, the Tech team is applying GRATEFUL to social network analysis. But that same technology could also be used for any number of security applications, such as identifying hackers trying to break into a network. And, eventually, the technology developed under GRATEFUL could find its way onto smaller, more efficient computers in unmanned aerial vehicles or worn by soldiers.
The team is currently one year into a potentially five-year effort. Bader says most of the work is still in the elementary stages, but the team is developing proofs of concept software. “Our goal is to create architecture-independent software that can run across multiple hardware platforms and still perform extremely well,” he says.
]]>The award funds a collaborative effort between Yahoo Labs and Chau, who will work with Hua Ouyang, a Yahoo research scientist, on new approaches to discover, analyze, and visualize potentially compromised accounts in Yahoo Mail.
Yahoo Labs is committed to forging strong alliances with top U.S. faculty members by collaborating on cutting edge research to advance web science.
The FREP award represents yet another milestone between Yahoo Labs and Georgia Tech research collaboration efforts. FREP is Yahoo Lab’s primary academic outreach initiative, which is designed to produce the highest quality scientific collaborations and outcomes, and this joint research effort between Chau and Yahoo will solve problems of mutual interest with measurable outcomes such as joint papers, advances in algorithm and systems research, and visualization and interaction design.
Georgia Tech recognizes Seth Tropper and Kim Capps-Tanaka from Yahoo Labs Academic Relations for their contributions to this new endeavor.
]]>]]>