The November survey of computing news of interest to ASCR was compiled by Jon Bashor (JBashor@lbl.gov) with news provided by Argonne, Fermi, Lawrence Berkeley, Lawrence Livermore, Los Alamos, Oak Ridge and Pacific Northwest national labs.  Contact information and links to additional information, where available, are included in each article.

Research News...
Sandia Releases State-of-the-Art 3D Fluids-DFT Application
Berkeley Unified Parallel C 2.6.0 and GASNet 1.10.0 Are Released
LBNL Team Wins Best Poster Award at SC07 Conference
Berkeley Lab Researchers Win Best Poster Award at IEEE VAST 2007
Researchers at ORNL Probe the Nature of Fleeting Nuclei
ASCR AMR Research Prompts New Software Tools for Compatible Discretizations
SNL Researcher Highlights ASCR Mesh Quality Research at Meshing Roundtable
Sandia Analysis of Stochastic Reaction Networks Finds Biomedical Applications

People...
Kathy Yelick Named New NERSC Division Director
ORNL's Zacharia Takes on VP Post at the University of Tennessee
ORNL Researcher's Student Awarded HPC Fellowship
Sandian Co-Chairs Society for Industrial and Applied Mathematics Annual Meeting
ORNL Researchers Chosen for Franco-American Conference
INCITE User Peter Bradley of Pratt & Whitney Gives MasterWorks Talk at SC07

Facilities/Infrastructure...
NERSC Accepts 20,000-Processor Cray XT4 for Production Use
ANL's Green Blue Gene/P Adds Muscle to Address Challenging Scientific Problems
ESnet, Internet2 Complete Coast-to-Coast Backbone for ESnet4
New Globus Incubator Project

Outreach...
ORNL-Sponsored Sixth Annual Day of Science Draws Record Attendance
Nashville Conference Explores Challenges of Petascale Computing
ORNL Highlights Accomplishments at SC07
Korean TV Crew Spends a Day at NERSC, LBNL
Lustre Workshop Held at ORNL's NCCS

Research News:
Sandia Releases State-of-the-Art 3D Fluids-DFT Application
Tramonto 2.1 is the first open source public release of software for the solution of modern density functional theories for inhomogeneous fluid systems (Fluids-DFTs).  Tramonto is highly scalable and is the first application to support general solution of Fluids-DFTs in 3D. Prior to the wide availability of Tramonto, systems were commonly studied in 1D.  Now with Tramonto, 3D results are possible for problems ranging from adsorption in simple porous media (planar, cylindrical or spherical confining surfaces) to surface forces and surface induced phase transitions, with fluids ranging from simple atomistic fluids to colloidal fluids to polymeric fluids.  Furthermore, Tramonto opens the door for investigation of systems with increased complexity including disordered porous media, biological membranes and molecules and nanocomposite materials.

The release of Tramonto 2.1 provides the science community access to new capabilities for systems of unprecedented complexity.  Tramonto version 2.1 is available as an Open Source package under the GNU Lesser General Public License (LGPL) and is available for download from the website http://software.sandia.gov/tramonto.
(Contact: Mike Heroux, maherou@sandia.gov)

Berkeley Unified Parallel C 2.6.0 and GASNet 1.10.0 Are Released
The Berkeley Unified Parallel C (UPC) team — a collaboration of Lawrence Berkeley National Laboratory (LBNL) and UC Berkeley — has released Berkeley UPC 2.6.0, which is available for download at http://upc.lbl.gov/download.  CDs of the release were distributed at the PGAS (Partitioned Global Address Space) booth at the SC07 conference this month.  UPC is an extension of the C programming language designed for high performance computing on large-scale parallel machines.  The language provides a uniform programming model for both shared and distributed memory hardware.  The UPC 2.6.0 release contains numerous improvements and fixes.  Bundled with this release is the new 1.10.0 release of the GASNet communication system, which is also available for separate download at http://gasnet.cs.berkeley.edu/.  Berkeley Lab members of the UPC/GASNet team include Kathy Yelick (PI) and Mike Welcome of NERSC and Dan Bonachea, Wei Chen, Jason Duell, Paul Hargrove, Costin Iancu and Rajesh Nishtala, all from the Computational Research Division.
(Contact: Ucilia Wang, UWang@lbl.gov)

Berkeley Lab Researchers Win Best Poster Award at IEEE VAST 2007
A group of Berkeley Lab researchers won the Best Poster Award at IEEE VAST 2007 (IEEE Symposium on Visual Analytics Science and Technology, Sacramento, CA, Oct. 30-Nov. 1, 2007).  Cecilia Aragon of CRD's Visualization Group and the NERSC Analytics Team; along with Stephen Bailey, Sarah Poon, Karl Runge and Rollin Thomas of the Physics Division were recognized for their poster "Sunfall: A Collaborative Visual Analytics System for Astrophysics," describing the first visual analytics system in production use at a major astrophysics project (the Nearby Supernova Factory).  A two-page abstract of the poster can be read at LINK and the poster itself can be viewed at this LINK.
(Contact: Ucilia Wang, UWang@lbl.gov)

Researchers at ORNL Probe the Nature of Fleeting Nuclei
Oak Ridge National Laboratory (ORNL) physicist David Dean and a team of researchers from the University of Tennessee (UT), Argonne National Laboratory and Iowa State University are using the National Center for Computational Sciences’ (NCCS’s) Cray XT4 Jaguar supercomputer to probe the nature of fleeting, unstable nuclei in an effort to better understand all nuclei, stable and unstable alike.  Dean discusses his work in the November 2007 issue of Physics Today, published by the American Institute of Physics.

Not all atomic nuclei are stable—just the ones that last.  Most of the elements in the universe are produced by exploding stars known as supernovas; and most of these new nuclei are unbalanced, being far too heavy on the protons or the neutrons and lasting only a tiny fraction of a second.  The team’s work has potential ranging from making nuclear power generation more stable and reliable to ensuring the effectiveness of America’s nuclear stockpile.
(Contact: Jayson Hines, hinesjb@ornl.gov)

ASCR AMR Research Prompts New Software Tools for Compatible Discretizations
Compatible discretizations translate partial differential equation (PDE) models into algebraic equations that inherit their key mathematical and physical properties.  Such discrete models are crucial for predictive simulations of complex physical phenomena of interest to DOE such as high-energy physics and electronic device modeling.  With the support of ASCR's Applied Mathematics Research (AMR) Program, Sandia researchers led by Pavel Bochev, in collaboration with M. Shashkov (Los Alamos National Laboratory) and university partners from Florida State University, have developed and analyzed a mathematical framework for compatible discretizations that unifies analysis of finite element, finite volume and finite difference methods.

Using this abstraction, Sandia researchers are designing interoperable software tools for compatible discretizations that offer access to a wide range of compatible finite element, finite volume and finite difference methods through a common programming interface.  A software package called Intrepid is currently under development that incorporates these ideas and the initial version of Intrepid will be available next year as part of the joint ASCR-NNSA Trilinos Project.  Intrepid takes advantage of the package architecture of Trilinos to provide interoperability with other Trilinos packages as well as future compatibility with PETSc and other scientific software tools and applications.  Sandia researchers have already used Intrepid to prototype new algorithms for the Sandia-NNSA Z-Machine, and future research will extend these methods and tools to a wide range of science applications.
(Contact: Pavel Bochev, pbboche@sandia.gov)

SNL Researcher Highlights ASCR Mesh Quality Research at Meshing Roundtable
Patrick Knupp recently presented an invited short course at the 16th International Meshing Roundtable that highlighted ASCR-funded applied mathematics research in mesh-quality optimization performed at Sandia National Laboratories.  Knupp's research on automatic computational methods for mesh quality improvement enables greater simulation accuracy and efficiency for scientific applications.  For example, Knupp and colleagues have discovered a new Target-Matrix paradigm that applies to the wide variety of mesh improvement problems encountered by applications such as mesh inversion, deforming domains, ALE rezone and mesh adaptation.  In prior mesh improvement algorithms, the collection of these problems could only be addressed through distinctly different algorithms requiring, in aggregate, more complex software infrastructure and higher maintenance costs.  This research is being deployed to the scientific user community through the general-purpose SciDAC-ITAPS Mesquite mesh-quality library.  Applications impacted thus far by ASCR mesh quality research, as implemented in Mesquite, include the Sandia Alegra/Z-pinch collaboration (enabling new ALE calculations that previously could not be done) and the CSAR Rocket Center (ensuring high quality tetrahedral meshes under deforming domains).
(Contact: Patrick Knupp, pknupp@sandia.gov)

Sandia Analysis of Stochastic Reaction Networks Finds Biomedical Applications
While chemical reactions on the macroscale are generally well described using ordinary differential equations, this does not hold on the nanoscale.  On this scale, reactions among small numbers of molecules behave more like Jump-Markov processes, with inherent stochastic fluctuations.  This variability often plays an important role in biosystems at the single-cell level, affecting processes such as gene transcription and cell signaling.  While many methods are available to simulate such networks, there are few tools for analyzing their dynamical behavior allowing for inherent noise and uncertainty.  A team of researchers at Sandia National Labs, led by Bert Debusschere and supported by the ASCR AMR program, is developing mathematical methods for analyzing such stochastic reaction networks, relying on spectral representations of random variables for sensitivity and predictability analysis and on Karhunen-Loève decompositions for reduced order modeling.  This work has led to a new collaboration with the research group of Allan Brasier at the University of Texas Medical Branch in Galveston, TX, to study the role of stochastic variability in immune system signaling pathways.
(Contact: Bert Debusschere, bjdebus@sandia.gov)

People:
Kathy Yelick Named New NERSC Division Director
Kathy Yelick, head of Lawrence Berkeley National Laboratory’s Future Technologies Group, a professor of computer science at UC Berkeley and an internationally recognized expert in developing methods to advance the use of supercomputers, has been named director of the NERSC Division at Berkeley Lab.  Yelick, who has been head of the Future Technologies Group since 2005, will officially assume her new job in January 2008.

Yelick has received a number of teaching and research awards and is the author or co-author of two books and more than 75 refereed technical papers on topics covering parallel applications, libraries, languages, compilers and architecture.  She earned her Ph.D. in computer science from MIT and has been a professor at UC Berkeley since 1991 with a joint research appointment at Berkeley Lab since 1996.  Read her longer biography at <http://www.nersc.gov/news/newsroom/yelick-bio.php>.

ORNL's Zacharia Takes on VP Post at the University of Tennessee
Thomas Zacharia, ORNL’s associate laboratory director for computing and computational sciences, was recently named Vice President for Science and Technology at the University of Tennessee (UT).  Zacharia will continue to lead the Computing and Computational Sciences directorate at ORNL.

UT Executive Vice President David Millhorn said he expects having Zacharia join UT’s senior leadership will strengthen the university’s goal of becoming a leader in high performance computing.  ORNL’s computational ties to the university go back decades but gained momentum with the establishment of the Joint Institute for Computational Sciences (JICS), which was the first of four new state-funded joint institutes and particularly with the recent award to UT of a National Science Foundation-funded supercomputing center, which will be housed at ORNL in JICS.
(Contact: Jayson Hines, hinesjb@ornl.gov)

Sandian Co-Chairs Society for Industrial and Applied Mathematics Annual Meeting Tamara Kolda of Sandia National Laboratories is co-chair for the 2008 SIAM Annual Meeting, to be held July 7-11, 2008 in San Diego, California.  Several of this year's meeting themes are relevant to the DOE, such as computational science and engineering, geosciences, data mining, imaging science and scientific software.  Featured DOE speakers include Karen Devine (SNL) and Anders Petersson (Lawrence Livermore National Laboratory).  Also, David Brown (LLNL) serves on the organizing committee.  For more information, see http://www.siam.org/meetings/an08/.
(Contact: Tamara Kolda tgkolda@sandia.gov)

ORNL Researcher's Student Awarded HPC Fellowship
Dr. Sudharshan Vazhkudai's student intern, Chao Wang, has been awarded the ACM/IEEE-CS HPC Ph.D. Fellowship.  The fellowship was awarded for Chao Wang's previous/planned work on HPC center storage and distributed storage research with Dr. Vazhkudai (of ORNL’s Computer Science and Mathematics Division) as well as work dealing with fault tolerance he conducted with his advisor, Frank Mueller, at North Carolina State University.  This is the first year of this extremely competitive program, honoring exceptional Ph.D. students throughout the world with the focus areas of High Performance Computing, Networking, Storage and Analysis.
(Contact: Jayson Hines, hinesjb@ornl.gov)

ORNL Researchers Chosen for Franco-American Conference
Two researchers at the NCCS recently took part in the Young Engineering Scientist Symposium in Washington, D.C.  Bronson Messer and Richard Mills were selected for their early career achievements and outstanding research in computational science.  The event, held October 22-24 at the French embassy, gathered 20 young (less than 10 years past their Ph.D.) scientists, 10 from France and 10 from the United States, to brainstorm together for possible collaborative research projects.

"Almost everyone got involved for some initial proposal for a research project," said Messer, whose proposals included projects for preconditioners for linear systems and solvers for magnetohydrodynamics. "It was great to learn about new solver technology, especially avenues that aren't being explored in the U.S."
(Contact: Jayson Hines, hinesjb@ornl.gov)

INCITE User Peter Bradley of Pratt & Whitney Gives MasterWorks Talk at SC07
At the SC07 Conference in Reno, Nevada, Peter Bradley of Pratt & Whitney delivered a MasterWorks presentation describing P&W's research efforts in aircraft engine combustor design and testing.  The critical role of high performance computing was discussed in support of CFD modeling and advanced simulation.  Peter Bradley described P&W's participation in the DOE's INCITE Program and the very successful collaborations P&W has had with Argonne's Leadership Computing Facility (ALCF) as part of the INCITE project efforts.

Facilities/Infrastructure:
NERSC Accepts 20,000-Processor Cray XT4 for Production Use
NERSC announced in October that it has accepted its 20,000-processor Cray XT4 supercomputer and put the system into production use.  With a top processing speed of more than 100 teraflops, the next-generation supercomputer will be used to advance a broad range of scientific research.  The computer was ranked ninth in the November 2007 TOP500 list.  The XT4 is named “Franklin” in honor of the first internationally recognized American scientist, Benjamin Franklin.

During negotiations to procure the system, NERSC and Cray mapped out a plan to install the Cray Linux Environment (CLE) on each of Franklin’s 9,672 nodes.  As a result of this partnership, NERSC became the first center with a production XT4 running CLE, an ultra-lightweight version of the standard Linux operating system, which makes the system easier to use, allowing users to more easily adapt their scientific applications from other architectures.  During extensive testing, about 300 different features and functions were tested and validated, making CLE more reliable with the same or better performance than previous XT operating systems.
(Contact: Ucilia Wang, uwang@lbl.gov)

ANL's Green Blue Gene/P Adds Muscle to Address Challenging Scientific Problems
This month, IBM and DOE's Argonne National Laboratory announced completion of a contract for a 445-teraflops Blue Gene/P system for the Argonne Leadership Computing Facility (ALCF).  As the ALCF's second major acquisition, this enhancement increases the system capability by a factor of five to 556 teraflops.  This additional capacity will accelerate the coming era of petascale computation in support of breakthrough science and engineering aimed at solving the nation's most challenging scientific problems.

Regarding pure computation power, the Blue Gene/P can carry out 445 million million calculations per second (445 teraflops is 445 followed by 12 zeroes).  Moreover, the Blue Gene/P systems consume a fraction of the power per teraflop required by similar systems built around commodity microprocessors.  This energy-efficient solution reduces power demands and lowers operating costs.  The ALCF-Blue Gene/P system pairs one of the world's fastest machines with advanced data management capabilities to meet the intense computing and data demands of petascale computing by leveraging the combined power of Argonne, IBM, Myricom and DataDirect Networks.
(Contact: Paul Davé, dave@alcf.anl.gov)

ESnet, Internet2 Complete Coast-to-Coast Backbone for ESnet4
Two of the nation’s leading research networks — DOE’s Energy Sciences Network (ESnet) and Internet2 — announced November 14 that they have completed five interconnected rings, each consisting of one or more 10-gigabit-per-second (Gbps) paths, that form a coast-to-coast network that is the backbone of DOE’s next-generation scientific network. Called ESnet4, the new network built by this partnership is a highly reliable, high capacity nationwide network that will greatly enhance the capabilities of scientists at national laboratories and universities across the country.  The network is managed by ESnet and operated across Internet2’s backbone infrastructure.

Steve Cotter, lnternet2 Director of Network Services and Deputy Operating Officer, praised ESnet “for your diligent efforts [for a] … very aggressive network build out and transition plan that included 570 individual tasks compressed into 261 days and completed over 1,350,000 million gigabit miles of optical networking capacity.”  The next steps will be to add a sixth ring connecting the northern and southern sections of the national network and to increase the capacity of all of the rings by adding 10 Gbps to the existing rings.  Once complete, these connections will increase both bandwidth and network reliability for the tens of thousands of DOE-supported researchers at national laboratories and universities around the country.  Read more at this LINK

New Globus Incubator Project
The Remote Application Virtualization Infrastructure (RAVI), being developed at Argonne National Laboratory as part of the SciDAC Center for Enabling Distributed Petascale Science (CEDPS), has been accepted as the newest Globus Incubator project.  The Incubator process was established two years ago as a means for including new projects and functionality as part of the software and tools in the Globus Toolkit.

The goal is to encourage groups to contribute their work while ensuring consistency and minimal duplication of projects.  RAVI is intended to provide GUI-based tools to guide the user through the process of identifying an application, mapping from strongly typed Web Services operations to application arguments, defining authentication and authorization requirements and deploying a service onto an execution site.
(Contact: Gail Pieper, pieper@mcs.anl.gov).

Outreach:
ORNL-Sponsored Sixth Annual Day of Science Draws Record Attendance
Over 1,200 students from 125 colleges and universities descended on the Knoxville Convention Center on Monday, October 29, for the Sixth Annual Day of Science.  The event, sponsored by the Department of Energy and Oak Ridge National Laboratory, was one of the largest DOE educational events ever held and was three times larger than any previous Day of Science, featuring speakers and exhibits from government laboratories and universities aimed at recruiting the next generation of American scientists. Participants were shown firsthand the latest developments in green energy, nanoscience, fusion and astrophysics, to name a few.  Among the participating ORNL departments were the Computer Science and Mathematics Division, the Computer Science and Engineering Division, the Information Technology Services Division, the Computing and Computational Sciences Directorate, the Research Alliance in Math and Science program and the National Center for Computational Sciences.
(Contact: Jayson Hines, hinesjb@ornl.gov)

Nashville Conference Explores Challenges of Petascale Computing
About 120 leading experts from academia, industry and government convened in Nashville in late September for the 2007 Fall Creek Falls Conference to explore the challenges of designing and running models and simulations on high-end computing systems performing at the level of petaflops, or a quadrillion floating point operations per second.  The conference, held annually since 2004, brought together users of high performance computing facilities for open scientific research, software developers and hardware vendors to discuss how to wring the most science from high-end facilities provided by the DOE Office of Science.  The event included individual presentations, panel discussions and posters organized around themes including climate change, energy and hardware and software including scientific visualization tools.
(Contact: Jayson Hines, hinesjb@ornl.gov)

ORNL Highlights Accomplishments at SC07
Oak Ridge National Laboratory showcased its leadership accomplishments at SC07, the premier international conference for high-performance computing, networking, storage and analysis.  ORNL's booth, titled "Leadership Computing for Transformational Science," featured all-electronic content, interactive kiosks and a large visualization wall which allowed attendees to learn more about the laboratory's research in alternative energy solutions, astrophysics, climate modeling and fusion energy.

Many of the lab's researchers helped to plan and conduct the conference.  ORNL's Becky Verastegui was the general chair and several other ORNL staff members served in key committee positions.  Besides co-organizing eight birds-of-a-feather meetings and three workshops, ORNL researchers presented papers at the conference and the lab's booth featured five talks on various aspects of high-performance computing.  SC07 was held November 10-16 in Reno, Nevada.

Korean TV Crew Spends a Day at NERSC, LBNL
On November 16, a television crew from YTN, a TV network in South Korea, spent a full day conducting interviews and getting shots of the NERSC machine room and LBNL site. The crew is working on a documentary about supercomputing.  Joining the visit were representatives from KISTI, the Korea Institute of Science and Technology Information. The program is scheduled to be broadcast in January.

Lustre Workshop Held at ORNL's NCCS
The National Center for Computational Sciences (NCCS) recently held a Lustre systems administration workshop at Oak Ridge National Laboratory.  Lustre is a highly scalable file system used on some of the world's fastest supercomputers, including the Jaguar system at the NCCS.  NCCS group leader for technology integration Shane Canon led the Oct. 15-16 workshop.  The workshop focused on a review of the Lustre file system and how it works and specific details of tools and tricks system administrators can use.  Canon said follow-up workshops are planned.
(Contact: Jayson Hines, hinesjb@ornl.gov)