CSMD

The Computer Science and Mathematics Division (CSMD) is ORNL's premier source of basic and applied research in high-performance computing, applied mathematics, and intelligent systems. Basic and applied research programs are focused on computational sciences, intelligent systems, and information technologies.

Our mission includes working on important national priorities with advanced computing systems, working cooperatively with U.S. Industry to enable efficient, cost-competitive design, and working with universities to enhance science education and scientific awareness. Our researchers are finding new ways to solve problems beyond the reach of most computers and are putting powerful software tools into the hands of students, teachers, government researchers, and industrial scientists.

News

CSMD's Nageswara Rao Wins R&D 100 Award (July 2014)
Diagnosis Using the Chaos of Computing Systems, or DUCCS, was developed at ORNL by Nageswara Rao.

DUCCS ultra-lightweight software quickly and nonintrusively detects a variety of hardware faults in processing units, accelerators, memory elements and interconnects of large-scale high-performance computing systems such as supercomputers, clusters and server farms. The software combines chaotic map theory with advanced CPUs and CPU systems to detect component faults in systems that handle large computational problems such as scientific computations, weather predictions and web data processing. DUCCS software provides critical diagnosis information that contributes to the resilience of computing systems in terms of error-free computations and sustained capacity.

The research was funded by the DOE's Mathematics of Complex, Distributed, Interconnected Systems Project, Applied Mathematics Program, Office of Advanced Scientific Computing Research.

 

Sreekanth Pannala Delivers ORNL Symposium Talk "High performance computing for accelerating sustainable transportation innovations" (May 2014)

Making significant gains in vehicle efficiencies involves understanding extremely complex interactions among the engine, battery, power electronics, materials, aerodynamics, and controls. Computational science is a growing field with the capacity to handle the exponential increase in the vehicle calibration space and simulate real-world scenarios. In his talk, Dr. Sreekanth Pannala detailed the computational efforts related to transportation and lay down a roadmap for using high performance computing to accelerate the design of highly efficient and cost-effective vehicles that meet regulatory, safety, and consumer demands.

Read the full abstract [here]. View the presentation [here].

Kate Evans' Work Recognized by Alma Mater (May 2014)

The work of CSMD researcher and Group Leader Kate Evans is currently being highlighted in an article in Haverford College's quarterly online magazine. The article "Investigating a Changing Climate" outlines work being carried out by Haverford graduates. In particular, the article touches on the work Kate has done with the Community Climate System Model which has been modeling global climate with increasing complexity and fidelity since 1983. Running on ORNL's Titan supercomputer, this model is able to calculate temperature, wind, rain, humidity, and other variables on a virtual grid covering the globe. These models can reproduce the past and predict how the climate will change in the coming decades and centuries.

Please go to http://www.mydigitalpublication.com/publication/?i=200366 to read the entire article.

This visualization of water vapor from a climate simulation was developed by the research group of Kate Evans using the Community Climate System Model, version 3. Monitoring water vapor enables climate scientists to improve their understanding of regional water cycles within a global context.-- Image created by Jamison Daniel

 

 

 

 

 

 

 

 

Patent Issued for "Coordinated Garbage Collection for RAID Array of Solid State Disks" (May 2014)

A team of researchers at ORNL invent a novel architecture for high performance non-volatile storage systems.

Non-volatile solid-state storage technologies such as Flash storage are increasingly used to accelerate I/O intensive workloads across a broad spectrum of domains from scientific computing to cloud computing and traditional enterprise IT. These storage technologies hold the promise of significantly improving the performance of a variety of workloads be they hadoop workloads, massive scale simulation outputs, to enterprise database workloads. While these technologies hold significant promise to accelerate these workloads, many solid-state storage technologies can suffer from significant performance degradation due to internal management tasks within the solid-state device. These internal management tasks, most notably garbage collection tasks used to reclaim free storage space after, can reduce performance by as much as 70 percent. While these internal management tasks occur infrequently on any one device, as organizations deploy multiple solid-state storage devices in common storage configurations such as RAID, the probability of performance degradation increases dramatically. To alleviate this, the ORNL team developed a novel architecture to coordinate these internal storage management tasks across a large pool of storage devices. This architecture significantly reduces the probability of productive I/O operations becoming degraded by these internal management tasks. For further information please refer to the full patent (patent number 8,713,268 B2) and the paper: David A. Dillow, Galen M. Shipman, Feiyi Wang, Sarp Oral, Junghee Lee, Youngjae Kim, "Coordinating Garbage Collectionfor Arrays of Solid-State Drives," IEEE Transactions on Computers, vol. 63, no. 4, pp. 888-901, April 2014, doi:10.1109/TC.2012.256

CSMD's Nagiza Samatova Wins Award (April 2014)

CSMD researcher Nagiza Samatova has been awarded the IEEE Computer Society's award Distinguished Contributions to Public Service in a Pre-College Environment for her outstanding teaching and mentoring of pre-college students in research projects.

Please see the complete award [here].

 

 

 

 

Postdoctoral Researcher Tianyu Jiang Awarded in "Best Dissertation" Competition

Tianyu Jiang, a postdoctoral researcher at Oak Ridge National Laboratory's Climate Change Science Institute, achieved 3rd place in the Chinese-American Oceanic and Atmospheric Association's "best dissertation" competition. Jiang, a member of CCSI's Earth System Modeling Group, received the award February 5 in Atlanta during the 94th annual meeting of the American Meteorological Society.

Jiang said what gave him the upper hand in the competition was a summer spent as a part of ORNL's Higher Education Research Experiences program, under the supervision of Kate Evans, leader of the ORNL Computational Earth Sciences Group and a member of CCSI. "My experience here was world class," said Jiang. "I had the chance to use high-performance computers and collaborate with the entire team to do cutting-edge work."

Jiang's thesis, "Understanding the Scale Interaction of Atmospheric Transient Disturbances and Its Coupling with the Hydrological Cycle over the Pacific-North American Regions," focused on the connection between climate over the North Pacific and extreme weather - freezing temperatures, high-impact precipitation, droughts, and other severe events. "Understanding extreme weather under a changing climate is becoming more and more challenging and urgent, not only because of the public's immediate need for weather information, but also the increasing concern from policymakers," said Jiang.
- by Justin Kaffka

Paul Kent was elected as a member at large in the American Physical Society, Division of Computational Physics (DCOMP).

Forrest Hoffman's paper "Causes and Implications of Persistent Atmospheric Carbon Dioxide Biases in Earth System Models" was recognized as the most downloaded paper from the Journal of Geophysical Research: Biogeosciences for the month of February. The complete author's list and information can be found [here].

Clayton Webster was elected Frontier of Science Fellow for the National Academy of Sciences.

A committee of National Academy members selects the Fellows from among the best young scientists in a broad range of science disciplines. Clayton was selected for his work in uncertainty quantification and he presented an overview of the field at the Kavli Frontiers of Science symposium, jointly organized by the Humboldt Foundation and the US National Academy of Sciences, April 4-7, 2014.

Kate Evans Appointed Group Leader

The Computer Science and Mathematics Division (CSMD) is pleased to announce the appointment of Dr. Kate Evans as the Group Leader for the Computational Earth Sciences Group within CSMD, effective January 1, 2014. Kate has been the interim group leader since February 2013, succeeding Danny McKenna, who has moved to the Graduate Education and University Partnership Division. Kate received her Ph.D. from Georgia Institute of Technology in Atmospheric Science and her B.S. from Haverford College in Physics. Kate will be responsible for leading and continuing to develop a top notch research capability in computational Earth science.

 

 

 

 

Bobby Sumpter New Deputy Director

Bobby G. Sumpter has accepted the position as Deputy Director of the Center for Nanophase Materials Sciences (CNMS).

Bobby received his Ph.D. in Physical Chemistry from Oklahoma State University in 1986. Following postdoctoral studies in Chemical Physics at Cornell University and in Polymer Chemistry at the University of Tennessee, Bobby joined the Chemistry Division at Oak Ridge National Laboratory as a staff member in the Polymer Science group. Bobby's research is focused on the fundamental understanding of self-assembly processes, interactions at interfaces, the structure and dynamics of molecular-based materials, and the physical, mechanical and electronic properties of nanoscale materials. His research uses a broad spectrum of materials theory and large-scale simulation approaches, including electronic structure and molecular dynamics. Working at the confluence of theory and experiment, he has authored over 300 scientific publications and several patents. He was named Corporate Fellow in 2013.

Bobby has been a staff member of the Center for Nanophase Materials Sciences since 2007 and the group leader for the Nanomaterials Theory Institute (NTI) at the CNMS since 2009. He is also the group leader of the Computational Chemical and Materials Sciences (CCMS) group in the Computer Science and Mathematics Division. Bobby will continue to serve as acting group leader for both the NTI and CCMS.

Clayton Webster Appointed Group Leader

The Computer Science and Mathematics Division (CSMD) is pleased to announce the appointment of Dr. Clayton Webster as the Group Leader for the Computational and Applied Mathematics Group within CSMD. Clayton received his Ph.D. from Florida State University in Applied and Computational Mathematics and his M.S. from McMaster University in Applied Mathematics. Clayton will be responsible for leading and continuing the development of research capability in mathematics.

Intern Roisin Langan wins best abstract award at student poster session

Roisin Langan, an intern at Oak Ridge National Laboratory (ORNL), spent last summer improving the ability of climate models to predict the variability and extremes of precipitation. With guidance from her mentors, Richard Archibald and Kate Evans of ORNL's Climate Change Science Institute, Langan analyzed data generated on the Oak Ridge Leadership Computing Facility's Titan supercomputer. Her project, titled "Stochastic Representation of Unresolved Processes in Climate Models," garnered attention labwide, winning the best abstract award at ORNL's Research Alliance in Math and Science (RAMS) banquet, a student poster session held August 8. Research in this field could result in more accurate warning systems for extreme events, such as flooding, droughts, and heat waves, and help stakeholders plan economic and humanitarian relief efforts. "This experience helped me gain invaluable networking channels, experience, and instruction in effective scientific communication," said Langan, a recent graduate of the University of California, Santa Barbara. Now an intern through ORNL's Nuclear Engineering Science Laboratory Synthesis program, she hopes to enter a graduate program in computational science in fall 2014. - Jennifer Brouner
Caption: Gil Weigand, director of Strategic Programs in the Computing and Computational Sciences Directorate, presents Roisin Langan with the best abstract award at the RAMS banquet.
Photo credit: Jason Richards, ORNL.

Science Highlights

Detection and Localization Algorithms for Networks of Radiation Detectors (April 2014)
Contact: Nagi Rao(raons@ornl.gov)

Challenges of Large-Scale Network Radiation Detection Algorithms: A network of detectors is a high-dimensional, non-linear, distributed system with a dynamic structure, wherein different detectors move in and out of sources proximities. Moreover, the detection and localization algorithms are driven by random radiation spectral counts that follow the high variance Poisson distribution. The performance optimization of the network-algorithms requires the ability to adapt not only to the dynamic structure and random measurements, but also to the background levels vary across the network.

CSMD Projects and Contributions: Over the past decade, a series of projects have been funded at CSMD by the Office of Naval Research (ONR), Department of Energy ASCR, Domestic Nuclear Detection Office (DNDO) and Defense Advanced Research Projects Agency (DARPA) in the area of network detection algorithms. The results from these projects led to a confluence of analytical results, efficient algorithms and test measurements that together meet several of the challenges of large-scale radiation detection networks. ORNL has developed a foundational theory (DOE Applied Mathematics Project) and experimental methods (ONR and DNDO projects) that overcome the limitations of current network fusion algorithms by using a novel network fusion approach, which leads to robust, dynamically optimizing detection and localization algorithms. The large-scale of DARPA Sigma project scenarios require further developments to handle the data volumes that stress the networks and algorithms, and large physical footprints that require judicious selection of subsets of data to be combined.

Our detection approach is novel and somewhat counter-intuitive: strategically-selected small subnets are applied to generate multiple source estimates, and their clusters are fused to yield a robust source estimate, which is then used to infer detection. The advantages of utilizing a network of detectors are illustrated in Figure 1, wherein a network of 12 detectors is deployed on the perimeter in a DNDO test. In this case, the network detects the source for the entire time as it moves across the network, whereas the non-networked detectors lose track of it in the interior. The main reason is that the network is able to continually localize the source whereas single detectors do it within their close proximity. Thus, the crux of our approach is to first run the localization algorithm using measurements from a select subnet, and then to infer detection using an adaptive, sharpened Sequential Probability Ratio Test. In essence, our approach combines all three elements, namely, the mathematical results, efficient localization algorithms and test datasets.

'SWEET SPOT' FOR SALTY WATER - ORNL-RPI simulations point to new material's potential in water purification (April 2014)

CSMD researcher and Group Leader Bobby Sumpter is part of a collaboration team from ORNL and Rensselaer Polytechnic Institute exploring the purification potential of a hybrid material called graphene oxide frameworks.

"This is basically sheets of oxidized graphene connected by specific chemical linkers from some of the oxidation sites," said Sumpter. "Because it's composed mainly of strongly bonded carbon, it doesn't decompose in water and has good mechanical properties. It's an exciting material with potential for numerous applications." Read the complete article [here].

Intermediate Frequency Atmospheric Disturbances: A Dynamical Bridge Connecting western U.S. Extreme Precipitation with East Asian Cold Surges in linking atmospheric weather extremes (April 2014)

Computational Earth Sciences Group members Tianyu Jiang and Kate Evans published the paper "Intermediate Frequency Atmospheric Disturbances: A Dynamical Bridge Connecting western U.S. Extreme Precipitation with East Asian Cold Surges in linking atmospheric weather extremes" in J. Geophysical Research: Atmospheres with collaborators at Georgia Tech and the University of North Dakota describing the improvement by global high-resolution climate models run on ORNL's Titan supercomputer in capturing the extreme weather events. The paper shows that by diagnosing observational data, over 80% of the extreme precipitation events over West Cost of the U.S. are associated with the landfall of atmospheric rivers (ARs), and the high-resolution model is able to reproduce this effect. The paper also illustrates how the East Asian cold air outbreaks influence the high-impact precipitation over the West Coast through the modulation of ARs, and that, unlike the lower resolution model, the high-resolution model correctly attributes the scale-interaction of the large scale atmospheric flow. Overall, this study provides an important scientific basis for extended forecasts, and demonstrates the need of global high-resolution models in analyzing climate prediction of variability and extremes.

The climatology of extreme moisture transport, termed 'atmospheric river' (AR) activity over the West Coast U.S. winter season (December-February) probability derived from the (a) lower resolution (~100km, T85) simulation, (b) observationally-based MERRA reanalysis data, and (c) higher-resolution (~37km, T341) simulation. The ratio of AR-induced extreme precipitation (99th percentile) is shown for the (d) T85 simulation, (e) MERRA reanalysis data, and (f) T341 simulation. (units: percent). The background of the figure shows an AR approaching the West Coast on Nov 30th 2012.

 

Multidisciplinary ORNL team discovers unexpected effect of heavy hydrogen in organic solar cells

Photovoltaic spray paint could coat the windows and walls of the future if scientists are successful in developing low-cost, flexible solar cells based on organic polymers. Scientists at the Department of Energy's Oak Ridge National Laboratory recently discovered an unanticipated factor in the performance of polymer-based solar devices that gives new insight on how these materials form and function.

"One of the dreams is to bring home some polymer paint from the hardware store, spray it on a window and make your own solar cell because it self-orders into a structure that can generate electricity," ORNL's David Geohegan said. "But right now there are many unknown things that happen when you spray it down and it dries. Changing the electrical property of a polymer also changes its structure when it dries, so understanding this process is one of our big science mysteries."

When ORNL scientists Kai Xiao and Kunlun Hong analyzed neutron scattering data obtained at the lab's Spallation Neutron Source to measure the structure of seemingly identical polymer-based solar devices, they stumbled upon a new piece to the scientific solar puzzle.

Read the rest of the story [here].

CSMD Researchers Speed Along Fusion Research

ORNL Team Members: Ed D'Azevedo, Pat Worley, and Dave Pugmire

CSMD researchers Ed D'Azevedo and Pat Worley were part of a team led by Princeton Plasma Physics Laboratory's C.S. Chang. The team increased the performance of its fusion XGC1 code fourfold on ORNL's Titan supercocmputer using Titan's GPUs and CPUs, compared to its previous CPU-only incarnation after a 6-month performance engineering period.

Read the full article [here].

Publication in March 2014 Cluster Computing

ORNL Team Members: Terry Jones and Stephen Poole

Computer Science Research Group member Terry Jones co-authored a paper describing recent successes from a collaboration between ORNL, the IOFSL Project, and the Vampir Project. The paper describes how the collaboration successfully improved the scaling capabilities of programming development tools. Programming development tools are a vital component for understanding the behavior of parallel applications. Event tracing is a principal ingredient to these tools, but new and serious challenges place event tracing at risk on extreme-scale machines. As the quantity of captured events increases with concurrency, the additional data can overload the parallel file system and perturb the application being observed. In this work, the authors presented a solution for event tracing on extreme-scale machines by increasing the maximum traced application size by a factor of 5x to more than 200,000 processes.

Thomas Ilsche, Joseph Schuchart, Joseph Cope, Dries Kimpe, Terry Jones, Andreas Knuepfer, Kamil Iskra, Robert Ross, Wolfgang E. Nagel, and Stephen W. Poole. Optimizing I/O Forwarding Techniques for Extreme-Scale Event Tracing. Cluster Computing. March 2014, Volume 17, Issue 1, pp 1-18. doi:10.1007/s10586-013-0272-9.

IBOBSU Website Enhancements: Progress Update

ORNL Team Members: Eric Lingerfelt, Michael Smith, Mitch Ferren, Monty Middlebrook

The National Isotope Development Center (NIDC) is the sole government source of stable and radio-isotope products for science, medicine, security, and applications. The NIDC manages the sales and distribution of these isotopes through the Isotope Business Office (IBO), which is located at ORNL. The Isotope Business Office Business Systems Upgrade (IBOBSU) project is an 18-month project begun in Summer 2013 that is focused on two separate tasks. The first task involves integrating the IBO's two internal business systems while the second task is comprised of major enhancements to the NIDC website at isotopes.gov and the NIDC Online Management Toolkit (OMT). Multiple milestones associated with the second task have been successfully achieved in the last three quarters. Website enhancements associated with these milestones include an online application for preferred customer status, a secure online quotation and ordering capability of stable isotopes from inventory for preferred customers, an online shopping cart feature for quote requests, and an online tool for preferred customer profile management. In addition to these deliverables, the 1.2 and 1.3 releases of the OMT provide software tools for NIDC staff that allow them to manage preferred customer accounts and process orders placed through the website.

This work is funded by the DOE's Office of Nuclear Physics under the Isotope Development and Production for Research and Applications (IDPRA) program.

Tamper-indicating Quantum Seals

CSMD, CSED, and MSSED researchers have developed a new technology for the technical verification of non-proliferation treaties. With support from the Defense Threat Reduction Agency (DTRA), the CSMD-led team has deployed quantum information concepts to detect when an intruder tampers with a sealed targets. This is a particular concern for treaty inspectors that need to confirm that the containment and surveillance of a special nuclear material has been uninterrupted. The novelty of the ORNL approach is to use quantum entanglement, a quantum mechanical feature that describes how two spatially disparate systems can exhibit seemingly strong correlations in their behaviors. The ORNL team has now leveraged those effects alongside the no-cloning principle to detect when an intruder is present, thus closing a vulnerability in existing tamper-indication technology. [more]

Software-defined Quantum Communication

CSMD researches have developed a new paradigm for leveraging the benefits of quantum communication. While quantum communication protocols like teleportation, entanglement swapping, and QKD are exciting possibilities for today's quantum communication engineers, these protocols are often exotic and unfamiliar to the end user. In addition, protocol implementation is often tightly coupled to the underlying physics, which makes tuning or reconfiguring the communication system difficult and costly. [more]

Adiabatic Quantum Programming

With the availability of first generation quantum computers, questions of programming and benchmarking are at the forefront of quantum computer science. A recent CMSD-led effort is helping address these questions by developing an integrated development environment for quantum computing. [more]

Nuclear Mass Dataset Dissemination and Analysis with the Nuclear Masses Toolkit and nuclearmasses.org

A knowledge of the masses of subatomic nuclei forms a crucial foundation for research in basic and applied nuclear science, as well as in astrophysics. New accelerator facilities and new detection systems have enabled researchers around the world to make more, and much more precise, nuclear mass measurements. In late 2012, a new Atomic Mass Evaluation was released, the first in a decade, that included all of this new information. However, the dissemination of these new masses (as an 850-page paper or as one enormous electronic table) has limited utility for researchers. Our Nuclear Masses Toolkit (NMT) provides the only online dissemination of these new masses whereby a ready comparison can be made with older masses and with the predictions of over 13 different theoretical mass models. [more]

Electronic control over molecular self-assembly and polymerization

CSMD and CNMS researchers Qing Li, Jonathan Owens, Chengbo Han, Bobby G. Sumpter, Wenchang Lu, Jerry Bernholc, Vincent Meunier, Petro Maksymovych, Miguel Fuentes-Cabrera, and Minghu Pan have demonstrated a non-thermal, electron-induced approach to the self-assembly of phenylacetylene molecules on gold that allows for a previously unachievable attachment of the molecules to the surface through the alkyne group and further controllable surface-coordinated linear polymerization of long-chain poly(phenylacetylenyl)s that are self-organized into a "circuit-board" pattern. [more]

New Version of C3

A new version of the Cluster Command Control (C3) tools has been released. The C3 tools are used as a core piece of the OSCAR cluster management suite, which has been updated to support the latest Ubuntu Linux distribution. These updated cluster tools are used internally by members of Computer Science Research to maintain group machines. They are also used stand-alone by a variety of users from industry, academia and laboratories. [more]

Edge-Edge Interactions in Stacked Graphene Nanoplatelets

CSMD researcher Bobby Sumpter was part of a team whose work on graphene platelets was published in the American Chemical Society's ACSNano Journal. [more]

Events

July 22, 2014 - Christos Kavouklis: Numerical Solution of the 3D Poisson Equation with the Method of Local Corrections