A publication of the Advanced Simulation & Computing Division, NA-121.2, NNSA Defense Programs
September 2008
NA-ASC-500-08—Issue 8Printable Version
(~1.1MB PDF)
The New Fiscal Year Cometh…
Editorial by Bob Meisner
For the past two years our budget has been determined by continuing resolution and it seems that we are facing that situation at least for the immediate future. There will be a continuing resolution up to the election and beyond and we are waiting to find out if the level of funding will be at 2008 levels or at the lower of the House and Senate marks. With a $66M difference between the marks, that decision will have a substantial impact on how we will be able to proceed.
Third Annual Modeling and Simulation Expo Held on Capitol Hill
On Tuesday July 15, 2008, the ASC Division exhibited for the second time in the third annual Modeling and Simulation (M&S) Expo sponsored by the National Training and Simulation Association (NTSA) and the Congressional M&S Caucus, co-chaired by Congressman J. Randy Forbes and Congressman Soloman P. Ortiz.
Earthquake Simulations Broaden Scientific Understanding
Using the ASC BlueGene/L supercomputer, Lawrence Livermore’s modern simulation tool, Wave Propagation Program (WPP), helps seismologists and earthquake engineers understand the ground motion hazard posed by earthquakes and other seismic events. The code computes ground motions for any seismic event in the San Francisco Bay Area region using a 3D geology/seismic wave-speed model of the area provided by the U.S. Geological Survey. Moderate earthquakes, such as the magnitude 5.6 quake that rocked the Calaveras Fault near the town of Alum Rock (east of San Jose) on October 30, 2007, can be used to validate earthquake simulations by testing the 3D model and allowing scientists to constrain uncertainties for large, less frequent earthquakes.
Multi-Agency Workshop Addresses Issues Facing Network Technologies
The Institute for Advanced Architectures and Algorithms (IAA), which is jointly operated by Sandia National Laboratories (SNL) and Oak Ridge National Laboratory (ORNL), hosted a highly successful workshop on advanced interconnection network technologies in San Jose, California on July 21 and 22. The workshop was organized by Scott Hemmert from Sandia and Jeff Vetter from Oak Ridge National Laboratory and sought to characterize the challenges facing interconnect technologies, looking toward building exascale supercomputers within the next decade.
Project Provides Thirteen Years of Industry-Shaping Deliverables
Over the past 13 years, the ASC File Systems and I/O project has provided products used by the ASC program directly, leveraged and shaped the HPC industry, and built and shaped the HPC file systems and I/O community.
The (former) ASCI Path Forward Program had two projects that were particularly noteworthy. The first Path Forward Global Parallel File System project was instrumental in guiding the design of the Panasas® File System, one of the most popular global parallel file systems in industry and used in production at Los Alamos. The second project was the guiding force behind the Lustre® file system, also a top parallel file system industry competitor and in use at Sandia and Lawrence Livermore national laboratories.
ASC’s Roadrunner Supercomputer Energy Efficient
In addition to breaking the petaFLOPS record, the Roadrunner supercomputer located in Los Alamos is #3 on the June 2008 Green500 list (http://www.green500.org/lists/2008/06/green500.php).
To illustrate the dramatic improvements in power efficiency, the first vector supercomputer installed at Los Alamos in 1976 needed 115 kW to deliver about 100 megaFLOPS on highly optimized matrix operations hand-coded in assembly language. To match Roadrunner's performance, ten million such machines would require 1,150,000 megaWatts of power — more than the entire electricity-generating capacity of the United States.
Red Storm Supercomputer Upgraded to 284 teraFLOPS
Located at Sandia, Red Storm is undergoing its second large-scale upgrade in four years of active duty. Sixty-five of the 135 compute cabinets are being upgraded from dual-core to quad-core Opteron processors by swapping dual-core modules for newer, quad-core boards. Processing power will increase from a theoretical maximum of 124 teraFLOPS to 284 teraFLOPS across 38,400 cores (12,960 nodes). To hold the computational results, roughly 1.5 petabytes of disk space are being deployed. A corresponding memory upgrade will provide 2 gigabytes of memory per core across the entire machine.
Maximizing Use of the ASC’s Purple Supercomputer
In the fall of 2005, the ASC Program appointed a team to formulate a governance model for allocating resources and scheduling the stockpile stewardship workload on ASC capability systems. Since October 2006, the ASC Purple capability system at Lawrence Livermore has successfully run as a national user facility using this model. The two primary objectives of the model are to ensure the capability system resources are allocated on a priority-driven basis according to program requirements and to use ASC capability systems for the large capability jobs for which they were designed and procured.
Collaborative Campaign Funding Aids Researchers to Develop New Modeling Capabilities
Researchers in Sandia’s ASC Verification and Validation Program, leveraging funding from NW Campaign 6 experiments, have developed new failure modeling capabilities within Sandia’s SIERRA Mechanics software framework. This new capability permits quasistatic and dynamic modeling of complex failure modes in composite materials, which are used in many weapon and non-weapon applications. This study employed Sandia’s DAKOTA software toolkit to generate over 800 SIERRA simulations, which were used to optimize parameters in the composite material failure models to match experimental test data, and, to perform sensitivity studies that identified the most critical parameters out of SIERRA’s 37-parameter material failure model.
Sandia System Software Researchers Receive SC08 Best Paper Nomination
A paper authored by Sandians Kurt Ferreira and Ron Brightwell together with Patrick Bridges from the University of New Mexico has been nominated for the Best Paper and Best Student Paper awards at the upcoming SC08 conference in Austin, Texas http://www.sc08.supercomputing.org/
The paper was one of four such papers selected from a total of 59 accepted papers. The paper entitled “Characterizing Application Sensitivity to OS Interference using Kernel-Level Noise Injection” is a detailed study performed on the ASC Red Storm machine of how operating system activity can impact the performance of parallel applications on very large-scale machines. The paper describes how the Catamount lightweight compute node operating system was extended to allow for creating various kinds of artificial noise to provide more insight into the important characteristics of applications and operating systems that influence performance and scalability.
Roadrunner Technical Seminar Videos Now Streaming on Web
Now available to the high-performance computing community are Roadrunner technical seminars presented by Computer, Computational, and Statistical Sciences (CCS) Division staff at Los Alamos National Laboratory.
These presentations about the first multi-core petaFLOPS speed computer, Roadrunner, were given March–June 2008 at Los Alamos. Using Flash Media Server, the hour-long sessions can now be seen in their entirety, including question and answer sessions, and can be used as tutorials as the adoption of similar systems increases. Visit the website at http://www.lanl.gov/roadrunner/rrseminars.shtml.
Sequoia Request for Proposals Approved, Proposals Accepted
The ASC Sequoia Supercomputer Request for Proposals (RFP) was approved for release by DOE/NNSA in July, and proposals were submitted by industry on August 21, 2008. The contract award should be announced by the end of the year. The RFP specified two machines: Sequoia with a mandatory 20-petaFLOPS peak with target delivery in 2011–2012 and Dawn with a mandatory peak of 0.5 petaFLOPS peak with target delivery in 2009. Bidders were required to run a series of benchmarks and include the results as part of their proposals and make projections of the performance of these benchmarks on the Dawn and Sequoia platforms.
“Designing scalable benchmarks is hard, but estimating the performance of a machine that is 20–50x faster than BlueGene/L and has not yet been designed poses even greater challenges,” said Mark Seager, assistant department head of advanced technologies.
Roadrunner Supercomputer Installed at Los Alamos
The first Roadrunner Phase 3 Connected Unit (CU) was delivered to Los Alamos on July 15, 2008. Shortly after delivery, IBM successfully kicked off a Linpack burn-in run on the first CU in the Metropolis Center for Modeling and Simulation.
Roadrunner Phase 3 System Stabilization Proposals Selected
Proposals have been selected in a competitive process to help with the Roadrunner Phase 3 system stabilization efforts once the machine is accepted at Los Alamos National Laboratory.
To advance science and programming knowledge, LANL has selected open-science and weapons-science projects.
ASC Salutes Matthew L. Leininger
As a deputy for Advanced Technology Projects within the ASC Program at Lawrence Livermore National Laboratory, Matt Leininger works at the center of advances in supercomputing that are enabling predictive scientific simulations.
“It’s exciting to see the direct impact advanced supercomputing architectures and predictive models are having on both scientific discovery and real-world problems such as stockpile stewardship,” said Matt.
DOE Privacy Disclaimer | Sandia Privacy Disclaimer | 2008-6301 W
Developed and maintained by Sandia National Laboratories for NA-121.2