NA-ASC-500-06—Issue 1

October 2006

The View from HQ

by Dimitri Kusnezov

As I write this, we don’t know the result of the upcoming elections, but whichever direction the elections go, we can expect the strong support of both sides of Congress to continue for the ASC program. We have a critical leadership role in national security and have demonstrated success in managing the program and delivering high-quality products. 

Read the full story…

Tri-Lab Test Suite Probes ASC Code Capability

LANL scientists achieved a new landmark with the first complete coverage of the Tri-Lab Verification Test Suite on 1-D, 2-D, and 3-D grids. The Tri-Lab Test Suite probes three aspects of an ASC code’s capability. The first inquiry examines the ability of the code to generate the correct answers to test problems with known solutions. The second probe scrutinizes the quality of those answers on uniform and nonuniform grids. The third probe quantitatively assesses how quickly the ASC code converges to the correct answer as the number of grid points is increased, and if that convergence rate is compatible with the expected theoretical maximum rate. Daily runs of the suite on a development code contribute to new releases of an ASC code being “born verified.” This work was part of the basis for two Defense Programs Awards of Excellence for the ASC Program in October 2006.

Read the full story…

Reconstruction Method Developed

A Sandia team (Stewart J. Mosso, Christopher J. Garasi, Richard R. Drake, and Allen C. Robinson) has developed and implemented a second-order accurate material interface reconstruction method in the ALEGRA High Energy Density Physics (HEDP) code. This method provides a higher fidelity approach for the simulation of material motion. The higher-order interface reconstruction algorithm is being applied to Rayleigh-Taylor unstable interfaces between conductive materials under rapid acceleration from very strong magnetic forces induced by large electrical currents. These interfaces are commonly encountered within simulations of experiments completed on Sandia’s Z machine. The simulation of material motion is particularly sensitive to the accuracy of the reconstruction method. Lower order methods produce more ejecta, which has a large effect on the material state, especially the compression/stagnation of the materials by magnetic forces. 

Winner of an R&D 100 Award, Sapphire Helps Find Useful Data

One of the great challenges researchers face today is extracting the information they need from enormous data sets. A Lawrence Livermore team, partly funded through ASC’s Pre- and Post-Processing Environment (PPPE) effort, has developed analysis algorithms allowing the exploration of large, complex, and multidimensional data sets. The technology has been dubbed Sapphire and recently captured one of seven R&D 100 Awards received by Lawrence Livermore from the trade journal R&D Magazine for being among the top 100 industrial innovations worldwide for 2006.

ASC Roadrunner Era Begins at LANL: Near-term capacity with a future hybrid accelerated option

The first components of the Roadrunner supercomputer arrived in Los Alamos on September 25, just two weeks after the contract with IBM was signed. The primary goals for the Roadrunner system are as follows:

• Providing a large-capacity computing resource for LANL weapons simulations
• Implementing optional upgrade to petascale hybrid accelerated architectures capable of supporting future LANL workload.
• Becoming a lead participant in the industry-wide path toward hybrid accelerated computing devices for HPC.

Milestone-Supported Study Shows Promising Results

A Sandia FY06 ASC Algorithms Level 2 milestone, “Algorithms for error-corrected reliability analysis in risk-informed design” was recently completed. It combines error estimation and adaptivity, uncertainty quantification, and probabilistic design capabilities for application to the analysis and design of bistable microelectromechanical (MEMS). The report reveals that through the use of error estimation and adaptive mesh refinement, solution verification can be performed in an automated and parameter-adaptive manner. The resulting uncertainty analysis and probabilistic design studies are shown to be more accurate, efficient, reliable, and convenient.

Sandia Achieves Turnkey Distance Visualization of Large Data ParaView

In September 2005, Sandia National Laboratories achieved a turnkey distance visualization milestone with the latest release of ParaView, an open source visualization tool. High energy density physics (HEDP) analysts in New Mexico visualized terabyte-sized data from ASC Purple in California. Though Purple has no specialized graphics hardware such as accelerated graphics cards, and the link itself is encrypted, the out-of-the-box solution of ParaView’s client/server architecture delivered results that delighted analysts. 

“Wow,” was one customer’s reaction, in comparing ParaView’s performance to other tools. ParaView delivered a 2x increase in performance for modest sized data (27 million unstructured cells), and was the only usable tool for an HEDP dataset consisting of 274 million unstructured cells.

Release of New Tools Improves Streamlining

On October 3, Sandia’s DART (Design Through Analysis Realization Team) group released an integrated suite of tools to benefit their customers within the analysis and nuclear weapons groups. This release, driven by past ASC milestones, provides analysts and designers a carefully integrated suite of tools that use a metadata layer to provide flexibility and agility. The tools are provided within a project management framework that also provides artifact tracking, pedigree, and archiving. The tools within the suite include modeling, geometric editing, decomposition, meshing, model management, material property management, workflow management, job submission, and post-processing.

Moab Software to Standardize Workload and Resources Across NNSA

The ASC Program has selected Cluster Resources, Inc.’s, Moab workload and resource management software as a standard for use across NNSA’s high-performance computing systems. The contract is the largest cluster and grid management contract in history. “Cluster Resources is honored to be selected,” said David Jackson, CEO of Cluster Resources, Inc. “There is no organization in the world which matches the technical expertise and scope of compute systems found at ASC in terms of scalability and architectural complexity.”

Contract Awarded for Tri-Lab Resource Management System

Cluster Resources, Inc. (CRI) is the winning bidder for a new tri-lab resource management system. This award marks the beginning of a multi-year collaboration between the tri-labs and CRI. Their resource management system, Moab, will first be deployed on the Atlas machine at Lawrence Livermore, which is a 44-teraFLOPS M&IC Peloton cluster. It will then be gradually migrated to other Lawrence Livermore systems. One of the key features of this solution is that users will continue to be able to use their familiar LCRM commands, through the implementation of wrappers and translators. For more information go to http://www.clusterresources.com

Los Alamos Develops V&V Course

During spring 2006, the ASC Verification and Validation (V&V) Program element at Los Alamos supported the development of a one-quarter graduate course on code verification, model validation, and the quantification of prediction uncertainty. The course was taught by François Hemez of the Applied Physics (“X”) Division through the Engineering Institute, a joint, multidisciplinary research and educational program between Los Alamos and the University of California, San Diego (UCSD). Course content includes activities for verifying the convergence of calculations, quantifying experimental and numerical uncertainty, eliciting expert knowledge, and validating simulation predictions. Fourteen graduate students passed the course, which was presented by video-conferencing and a video-streaming-enabled website hosted by the UCSD Jacobs School of Engineering. Student evaluation of the course was quite positive, and there are plans to expand this activity. Beyond its programmatic relevance for developing a predictive capability for the DOE Complex, the course serves the ancillary function of training and recruiting the next generation.

Ten-Year University Alliances Program Nears End

Annual site reviews were held during August and September 2006 of the five ASC Strategic Alliances Program (ASAP) centers of excellence: the California Institute of Technology, University of Chicago, University of Illinois at Urbana-Champaign, Stanford University, and University of Utah. The reviews were conducted by the associated Tri-lab Sponsoring Team (TST) for each Center and the Alliance Strategy Team (AST). The TST Leads led the reviews.

Follow-On University Program, Predictive Science Academic Alliance Program (PSAAP) Initiated

Following in the footsteps of the soon-to-conclude ASAP, the newly authorized Predictive Science Academic Alliance Program (PSAAP) has been initiated. Authorized by the ASC Execs, the primary goal of the PSAAP is to establish validated, large-scale, multidisciplinary, simulation-based “Predictive Science” as a major academic and applied research program.

A bidders meeting was held at the Dallas-Fort Worth Airport on May 16 and 17, 2006, and a Request for Interest (RFI) was released on June 15, 2006. There were 34 responses to the RFI received on July 17, 2006. Responses covered broad areas of scientific domains. A review of the RFI responses was completed on August 10, 2006, by a panel of experts from the three defense laboratories, other DOE national laboratories, federal agencies, and academia. Eighteen respondents have been encouraged to submit full proposals. The current plan is to release a Request for Proposals (RFP) in November 2006. The goal is to place PSAAP subcontracts by October 1, 2007. For more information, visit http://www.sandia.gov/NNSA/ASC/univ/psaap.html

ASCI Red Retired

ASCI Red was the first computer system to achieve more than 1 teraFLOP on the LINPACK benchmark, claiming the number one spot on the TOP500 list of major supercomputer systems in the world in June 1997. It was declared excess property and retired in July, 2006.

The disk subsystem was sanitized via physical destruction of the recording media, and the components of the system were designated scrap and physically rendered unusable. A ceremony in June marked the retirement/decommissioning the ninth anniversary of the commissioning of the system.

Read the full story…

ASCI White Retired

As reported in the October issue of the NNSA newsletter, ASCI White, the supercomputer at Lawrence Livermore, was shut down in July of this year.

Read the story on page 3 of the October NNSA Newsletter [PDF 1.2 MB] …

Recently  Completed:

• The major Tri-lab runs required for the Q1FY07 Purple, 100-TF Environment L1 milestone, which is led by Terri Quinn (LLNL), Rand Rheinheimer (LANL), and Judy Sturtevant (SNL), have been completed.
• The first set of Capability Computing Campaign (CCC) runs on Purple began on schedule on October 16, 2006.
• The LANL Roadrunner system is the first ASC platform to go through the DOE Order 413.3A process for “Program and Project Management for the Acquisition of Capital Assets.”
• 2006 Reviews for the ASC L1 Centers at Caltech, the University of Chicago, the University of Illinois, Stanford, and the University of Utah were completed in August and September. (See the article in the ASCeNews for more details.)

ASC Salutes

Editor’s note: Each quarter, the ASC Program will feature the outstanding contributions of one of its numerous tri-lab scientists, engineers, and administrators. This month, we proudly present Dr. Anthony Giunta.

Dr. Anthony (Tony) Giunta is a technical staff member in the Validation and Uncertainty Quantification Department within Sandia National Laboratories’ Engineering Sciences Center. For seven years, Dr. Giunta has been involved in a wide range of ASC-funded projects that span the research, development, and application spectrum.

Read the full story…

Photo Album

Photos from the ASC Executives Offsite Meeting in Vail, Colorado, October 11 and 12, 2006.

DOE Privacy Disclaimer | Sandia Privacy Disclaimer SAND 2006-6930W

ASCeNews Archive | Contact Us

Developed and maintained by Sandia National Laboratories for NA.114