A publication of the Advanced Simulation & Computing Division, NA-121.2, NNSA Defense Programs
June 2009
NA-ASC-500-09—Issue 11 Printable Version
(~1.5MB PDF)
The Meisner Minute
Editorial by Bob Meisner
I’d like to take a moment in this quarter’s issue to congratulate Lawrence Livermore National laboratory (LLNL) for their success in dedicating the National Ignition Facility (NIF) and the early delivery of the Dawn supercomputer. What a great day for simulation at the forefront of science! The dedication of both these capabilities in the same week is symbolic of not only the effort required to do modern science, but also that it’s happening at our laboratories.
“Dawn” of a New Era for Computing Partnership
The partnership that has dominated high-performance computing (HPC) for more than a decade was celebrated last week in a ceremony dedicating the Advanced Simulation & Computing (ASC ) Program’s latest supercomputer, Dawn.
The dedication in the Terascale Simulation Facility, at Lawrence Livermore National Laboratory (LLNL), was attended by representatives from the National Nuclear Security Administration (NNSA), the three national weapons laboratories, Lawrence Berkeley National Laboratory, and IBM.
Lawrence Livermore Hosts Predictive Science Panel; Expanded Charter Focuses on Improved Predictivity
The twice-a-year review of the ASC Program by the Predictive Science Panel (PSP) was held at Lawrence Livermore National Laboratory March 31 through April 2 to evaluate progress toward the goal of a credible predictive capability. Although the PSP has been evaluating ASC for years, this is the first time the meeting was held with the new expanded charter—to review what experimental data are being acquired and how they are being integrated into improved predictivity of the codes. As a result, the PSP has an expanded purview that now encompasses both ASC and its connection with the Science Campaigns. The original charter was more directly focused on ASC alone.
Los Alamos (LANL) Hosts 2009 Weapons Science Capability Review
The third annual meeting of the Weapons Science Capability Review was held this Spring at Los Alamos National Laboratory (LANL) in which the Weapons Physics Directorate hosted an external committee of reviewers to assess the quality of science, technology, and engineering within capabilities supporting weapons science. The ASC Program was well represented at this three-day meeting.
New RAMSES Simulation Software Tested to Assess Its Predictive Capability
In a joint ASC and Campaign 7 activity, a first-of-its-kind validation study was performed to assess the predictive capability of new features in Sandia’s RAMSES software. These new RAMSES capabilities were added to better assess weapon electrical circuit vulnerability to hostile X-ray and gamma-ray radiation environments. This study focused on electrical circuits constructed utilizing Sandia’s CMOS7 (complementary metal-oxide semiconductor) manufacturing process, which was used in the W76-1 life extension program and may be used in future weapon applications.
Sandia’s ASC Program Engages NASA in Risk-Informed Decision-Making Analysis
In late March 2009, a Sandia National Laboratories team visited the NASA Langley Research Center for a technical exchange on a variety of topics related to computational simulations and their use in risk-informed decision making. For the initial two days of the meeting, Sandia staff delivered a tutorial on sensitivity analysis and uncertainty quantification (UQ) methods, which was attended by 70 NASA staff, managers, and resident graduate students. These training materials were developed under the Verification and Validation (V&V) program element of Sandia’s ASC Program.
One of the discussions during this exchange centered on NASA’s recently published standard practice guide on the use of computational simulation tools in high-consequence applications. Over the past several years, representatives from the NNSA tri-laboratory community contributed to this document via technical discussions with NASA technical staff and managers. Currently, discussions are under way between Sandia and NASA staff to determine how some of Sandia’s expertise in UQ methods can be leveraged by NASA teams involved in aerodynamic performance assessments of new spacecraft launch vehicles.
Enhanced Verification Test Suite Completed for Physics Simulation Codes
A tri-lab team has completed a multiyear effort to identify and develop verification test problems to assess the numerical performance of models and algorithms implemented in ASC codes. The purpose of the verification analysis is to demonstrate whether the numerical results of the discretization algorithms in physics and engineering simulation codes provide correct solutions of the corresponding continuum equations.
ASC’s Seminars at Sandia on Verification, Validation, and Uncertainty Quantification Exceed Attendance Expectations
As of April 2009, the educational outreach efforts of the Verification and Validation (V&V) program element in the Sandia National Laboratories ASC Program have hit an attendance milestone: 1,000+ participants since the start of the educational activities in August 2007. These seminars and workshops have covered a broad array of topics related to computational simulations such as V&V, uncertainty quantification, mathematical optimization, and error estimation. Also included were related topics such as experimental test data uncertainty analysis and quantification of margins and uncertainties (QMU).
To date, there have been 35 seminars, workshops, and training sessions. Attendees have come from across Sandia (30+ centers and 170+ departments) including on-site and video-streamed sessions to Sandia personnel in Livermore and Carlsbad. In addition, a smaller number (~100) of external attendees have come from the NNSA Sandia Site Office, Los Alamos, Lawrence Livermore, the UK’s Atomic Weapons Establishment, NASA Langley Research Center, and the Air Force Research Laboratory. These ASC V&V educational activities will continue in FY09 and beyond. In addition, discussions are under way regarding possible integration with formal Sandia training and educational programs.
New Project Team Guides Code Developers to Fully Exploit Sequoia
The massive size and capability of the 2011–2012 Sequoia system presents significant challenges to users and staff.
With the 500-teraFLOPS ASC Dawn system on the computer room floor at Lawrence Livermore National Laboratory, preparation is under way to develop software applications scaled to run on Dawn’s successor, Sequoia. Sequoia will be capable of achieving performance levels 20 to 50 times that of BlueGene/L with materials science codes, and application code teams will need guidance to succeed in fully exploiting Sequoia’s capability. The Sequoia Scalable Applications Preparation Project (SSAPP) has been initiated by Livermore Computing (LC) to develop and provide the needed guidance for code teams.
New Scientific Insights on Magnetic Reconnection from Roadrunner Open Science Runs
As mentioned in the March 2009 issue of the ASC eNews, Los Alamos is conducting open-science runs on the Roadrunner petascale supercomputer. One of the science topics under investigation concerns magnetic reconnection, a basic plasma process in which magnetic field energy is rapidly converted into plasma kinetic energy. Understanding these basic structural features will allow researchers to better predict the dissipation rate and time dependence of reconnection for a variety of applications. Initial simulations of magnetic reconnection using the particle-in-cell code, VPIC, are leading to new scientific insights into the influence of plasma instabilities on the 3D evolution of reconnection layers. Go to http://www.lanl.gov/roadrunner/rropenscienceabstracts.shtml to see the abstract for this project “Three-Dimensional Dynamics of Magnetic Reconnection in Space and Laboratory Plasmas.” Principal Investigator, Bill Daughton, from the Applied Physics Division at Los Alamos is leading this project.
LANL Has Golden Opportunity to Prototype New Storage Archive
As archiving requirements for high performance computing (HPC) data grow while resources to maintain specialized archive systems stay the same or decline, a new approach to archiving may be needed. Newly available commercial storage technologies have opened the possibility to prototype an alternative solution to archiving data while minimizing ASC-funded code development and maintenance.
Los Alamos Tool Could Improve Checkpoint Bandwidth
The current method for massively parallel applications to protect themselves from component failure is through periodic checkpointing — a process in which applications save their state to persistent storage. Protection from component failure is increasingly important as systems grow in size and the number of components increases. Los Alamos National Laboratory is implementing a tool that will dramatically improve checkpoint performance for many applications.
Sandia’s DART Workbench Now in Broad Production Use for Model Management
In May 2009 Sandia National Laboratories released version 1.1 of the Design Through Analysis Realization Team (DART) Workbench —DWB, completing a three month “hardening” effort aimed at addressing all critical user issues identified after the initial release in early 2009. Over 200 individuals use a combination of DWB and ESAW (Engineering Sciences Analysis Workbench). New capabilities include fast file uploads, Python scripting, improved integration with Dakota (Jaguar UI), improved integration with Sandia's SIERRA codes (work in progress), job submission support for the new Tri-Lab Capacity Cluster (TLCC) machines, support for client-only mode, and multi-point constraint support for building the Salinas software input decks. Additionally, all known critical and high-priority bugs were addressed as part of this release.
These tools impact the user community via improved information management, with focus on management of model and simulation processes and data in support of computational simulation. They also act as a communication platform for the weapon engineering groups and directly support the weapon safety and reliability technical bases by acting as a repository for full system models, and associated analysis and test work.
Efforts are under way to provide these tools outside Sandia.
ASC Salutes Bob Little
Hardly a newcomer to ASC, Robert C. “Bob” Little, from Los Alamos National Laboratory, was selected this spring as Program Manager for the Physics and Engineering Models (PEM) p rogram e lement within ASC. The PEM program develops and implements the theory and models that are incorporated into simulation codes. As well as working especially closely with the Integrated Codes and V&V program elements, the PEM program plays an essential role in working with the experimental campaigns to help motivate and interpret measurements and to utilize measured results to improve models and data bases. Also, high-performance computing simulations sponsored by the PEM program on platforms such as Roadrunner and BG/L are providing exciting insights into the physical processes important to the nuclear weapons (NW) program as well as other areas of interest to the laboratories and the Department of Energy (DOE).
Brian Carnes Begins Position as Lawrence Livermore’s ASC Deputy Program Leader
Brian Carnes assumed the role of ASC Deputy Program Leader for Lawrence Livermore National Laboratory (LLNL) on April 1, 2009, replacing retiring ASC Deputy Lynn Kissel. Brian also will continue to manage the Multiprogrammatic and Institutional Computing (M&IC) program at LLNL as he has for the past six years. With more than 24 years at LLNL, Brian has spent most of that time in Livermore Computing (LC), serving in various functional and programmatic leadership roles.
Lynn Kissel Retires after 20 Years of Service
At the end of June, ASC Deputy Program Leader Lynn Kissel retires from Lawrence Livermore National Laboratory (LLNL) to pursue his love of vintage cars. Lynn has been with ASC for 10 years and for the past few months has been showing new ASC Deputy Brian Carnes “the ropes.”
Lynn first joined LLNL and the Nuclear Test Program in 1977 after receiving his Ph.D. in physics from the University of Pittsburgh. After spending the 1980s at Sandia Laboratories in Albuquerque and returning to the Physics Directorate at LLNL in 1991, he joined Livermore Computing and the Computation Directorate in 1998.
“It’s more than appropriate to thank Lynn Kissel for his service as ASC Deputy,” said ASC Program Leader Michel McCoy. “Lynn’s integrity, diligence, intelligence, gentle humor, and above all, his abiding loyalty to ASC and to me personally will not be forgotten. He will be missed.”
NY Times Extols Hyperion Technology
The New York Times featured an article about the Hyperion testbed (an advanced technology cluster testbed) in its Technology section earlier this month. According to the article, “Intel, Dell, Cisco Systems, Sun Microsystems, and others have worked with Lawrence Livermore to create something called the Hyperion testbed. In essence, it’s a supercomputer made of thousands of machines dedicated to pushing open source software to its limits and then refining the software for use on future machines built at various national labs.”
Mark Seager, assistant department head for advanced technologies at Lawrence Livermore National Laboratory, provided the keynote address entitled “The Challenges and Rewards of Petascale Clusters” at the High Performance Computing Asia and Asia–Pacific Advanced Network 2009 Conference in Kaohsiung, Tawain, in March.
“The conference had very good attendance from the Pacific Rim countries,” Mark said.“Many of the talks dealt with open science and the impact of global warming, and the technical topics covered a large array of issues, like supercomputing covers. There was also a vendor display area with about 30 vendor booths.”
DOE Privacy Disclaimer | Sandia Privacy Disclaimer | 2009-4030 W
Developed and maintained by Sandia National Laboratories for NA-121.2