A publication of the Office of Advanced Simulation & Computing, NNSA Defense Programs

NA-ASC-500-07—Issue 4

September 2007

The View from HQ

by Dimitri Kusnezov

Perceptions are not always reality

“Why does Dimitri hate our Lab?” This question was posed to a colleague recently—and passed on to me. My first response was surprise, which quickly turned to introspection. I do not think of the labs in such a context; rather, I think of them as dynamic entities. I see the labs through the people, through the trust and professional relations that develop over time and through their ability to work with us and deliver on commitments. While I might claim the question is ill-posed, that does not address the real issue: How does such a perception arise?

Complete story…

ASC Participates in Second Annual Modeling and Simulation Exhibition on Capitol Hill

The ASC Program recently participated in the second annual Modeling and Simulation Exhibition on Capitol Hill in Washington, DC. The exhibition was hosted by the Congressional Modeling and Simulation Caucus, which aims to promote the potential impact of modeling and simulation technologies on the nation’s military and defense capabilities, as well as on the broader economy. Currently, the Caucus is very focused on the use of simulation in military training and preparedness, and the exhibition strongly reflected these types of applications. The ASC Program was a unique and broadening contributor to the exhibition of simulation capabilities, with an emphasis on predictive science and high-performance computing.

World’s Fastest Supercomputer Delivers Breakthrough Science Simulations for the NNSA’s Nuclear Weapons Program

The BlueGene/L supercomputer at Lawrence Livermore National Laboratory topped the list of the world’s fastest computers for the sixth time, according to the new Top500 list released Wednesday, June 27, at the International Supercomputing Conference in Dresden, Germany.

Meeting Predictivity Milestones with High-Performance Computing

High-performance computing (HPC) is needed to support the Predictivity milestones for certification. In particular, HPC can be used to eliminate knobs by solving the many microphysics issues in relevant regimes and over the relevant range of spatial scales. This is called Science@Scale.

V&V Analysis Conducted of Large Simulation Results Using Scalable Python Scripting Capability

Verification of complex, large results from Sandia's simulation codes is crucial to understanding the science behind the results. Sandia’s scalable python scripting capability, released in ParaView 3.0, enables flexible, advanced V&V analysis of large data within a familiar python scripting environment. Using this capability, analysts write small modules in python, addressing specific solutions, or codes. The scalable python capability manages the rest (data distribution, load balancing, parallel python execution), making it simpler than ever for analysts to work on large data analysis problems.

Complete story …

Improving Supercomputer Reliability via Data Mining

Computer logs often provide critical information about malfunction or misuse, but finding and correlating the clues interspersed among millions of lines of time-stamped text messages generated by supercomputers is in itself a challenging data mining task. The Sisyphus toolkit is the result of four years of research and development on how to efficiently find the important nuggets of information in supercomputer logs. Now in production use on Red Storm, it has automatically detected—and more importantly isolated—a wide range of problems including failures (disks, I/O controllers, network interfaces, power supplies, and memory), misconfigurations (BIOS, RAID controller, system software, and inconsistent versions), and problematic user behavior (unbalanced RAID stripe usage, inappropriate remote monitoring). This has enabled focused proactive and reactive responses by system administrators, thus increasing system reliability.

New Model for Tantalum Developed under ASC Multiscale Modeling Effort

More than 10 years ago, an effort began to develop improved, physically based models of material constitutive behavior for use in integrated multiphysics codes. The challenge has been that the models are needed for programmatic applications in regimes of pressure, strain-rate, and temperature, where relevant experimental data are scarce or nearly impossible to obtain. Building upon the ASCI paradigm at the time, a multiscale modeling strategy was developed to bring together and leverage material modeling activities from a variety of length scales. This technique uses new and improved computational simulations involving detailed physics to enable the development of better macroscopic models for the integrated codes.

New Interface Reconstruction Method for Multi-Fluid Flows Developed as Part of LANL ASC Strategic Research

Los Alamos researchers Rao Garimella, Sam Schofield, and Marianne Francois, along with Raphael Loubere from Universite Paul-Sabatier, France, have developed a new interface reconstruction method for multi-fluid flows that is completely material-order independent. This is in contrast to traditional interface reconstruction methods, which produce very different results depending on the order in which materials are processed, leading to errors in advection of the materials.

New Conservative Remap Method in FLAG

Los Alamos mathematical modeling and analysis researchers Sam Schofield and Rao Garimella have implemented a new, prototype conservative remap method in the FLAG Arbitrary-Lagrangian-Eulerian (ALE) code that is based on exact intersections of the old mesh with the relaxed mesh. In ALE calculations, the mesh moves and distorts with time. When the mesh becomes tangled or of poor quality, it is smoothed, the solution values are mapped, in a conservative manner, from the old to the smoothed mesh, and the simulation continues. In the FLAG code (LANL ASC Code Project B), the existing remap method is based on approximating the changes in the values in the mesh cells between the new and old meshes. Because of accuracy constraints, it is limited to small displacements.

Progress on Simulating Wire-Array Z Pinch Implosions

A Sandia team has implemented a mass inflow boundary condition into the ALEGRA High Energy Density Physics code, which has allowed researchers to make significant progress in the study of Z pinches. A wire-array Z pinch is an annular array of tiny metal wires (~10 micron diameter) which, when pulsed with current, is subject to a magnetic force that implodes the pinch. The resulting stagnation and thermalization of the pinch on the cylindrical axis of symmetry yields a copious amount of x-rays, thus making the Z pinch an attractive radiation source. The first 50 to 80% of the Z pinch lifetime is described by a mass ablation phase, during which stationary wire cores cook off material that is subsequently swept towards axis by the magnetic field, thus resulting in a radial redistribution of mass. The ablation phase is difficult to simulate because it is both fully 3-dimensional (3D) in nature and requires very high resolution to model the small diameter wires.

Moab Rollout at Sandia

The Moab workload manager from Cluster Resources, Inc., was adopted by ASC as a Tri-Lab standard job scheduler last August. The job scheduler manages user’s compute jobs, allocates resources, and gathers accounting data. Users would have a common set of job management tools, systems administrators would be able to share expertise, and management would gain a common set of reporting metrics. The initial goal of the Tri-Lab Workload Management Board was to ensure a uniform job scheduling environment across all new ASC platforms. However, with a site-wide license in place, all three labs are moving rapidly to deploy the Moab scheduler on existing systems too.

LANL Roadrunner Base Capacity System Running ASC Codes

The Roadrunner project completed an ASC Level 2 milestone in June 2007, a demonstration of Roadrunner’s integration into the LANL classified computing system and success in running ASC codes. These milestone criteria were outlined in the FY07 ASC Implementation Plan. The system had been accredited for classified computing in May 2007.

Complete story …

LANL Roadrunner Procurement Team Wins Small Team Distinguished Performance Award

LANL’s Director Anastasio announced that the Roadrunner procurement team had been honored with a Small Team Distinguished Performance Award. Congratulations to team members Manuel Vigil and Ray Miller, High-Performance Computing-Division Office (HPC-DO); Diana Little, Computing Operations and Support (HPC-2); and Bart Burson, Acquisition Services Management-Purchasing (ASM-PUR). This team provided the leadership for the technical planning, procurement, and project management for the Roadrunner Platform.

Left: Director Anastasio presenting an individual distinguished performance award last year.

Sandia CSRI Workshop on Mathematical Methods for Verification and Validation

The Computer Science Research Institute (CSRI) of Sandia National Laboratories held a very successful workshop on Mathematical Methods for Verification and Validation (V&V), from August 14 to 16, 2007, at the Hyatt Regency Tamaya Resort. The workshop was chaired by Clayton Webster, the FY08 John von Neumann Fellow, and co-organized by fellow Sandians Scott Collis, Tim Trucano, and David Womble, as well as Prof. Max Gunzburger from the School of Computational Science at Florida State University. The conference Web site can be found at http://www.cs.sandia.gov/CSRI/Workshops/2007/MMVV/.

Recent Publication Highlights Science-Based Prediction at LANL

“Science-Based Prediction at LANL,” an article about how LANL scientists have taken on the challenge of building computational tools to examine complex physical processes by using innovations that extend beyond the classic scientific method, was published in the SciDAC Review, Issue 4, Summer 2007, p. 33, http://www.scidacreview.org/0702/html/hardware.html.

ASC Salutes

Dr. Michael (Mike) Heroux is a Distinguished Member of the Technical Staff in the Scalable Algorithms Department at Sandia National Laboratories. His research interests include the solution of challenging engineering and science problems, especially the solution of large-scale implicit problems and the development of robust preconditioners. His contributions in the past nine years have led to breakthroughs in large-scale circuit modeling, non-Newtonian fluid simulation, and classical density functional theories. As part of this work, Dr. Heroux has also been the primary developer of numerous linear algebra packages that are used by hundreds of other solver and application developers and are available as part of the Trilinos project. These tools, especially the Epetra package of scalable linear algebra classes, have been used within the ASC Program to develop scalable solvers across every major Sandia ASC-funded application.

Complete story…

DOE Privacy Disclaimer | Sandia Privacy Disclaimer |Contact Us | ASC eNews Archive

Developed and maintained by Sandia National Laboratories for NA-121.2