InTheLoop | 12.06.2010
December 6, 2010
NERSC’s Franklin Earns Second Place on Graph 500 List
Franklin, a Cray XT4 at the National Energy Research Scientific Computing Center, was ranked No. 2 on the first Graph 500 List, which was unveiled November 17 at the SC10 conference. The list ranks supercomputers based on their performance on data-intensive applications and thus complements the TOP500 list, which is based on the LINPACK benchmark. Read more.
CRD Scientists Get Allocations on Nation’s Fastest Computer
Scientists in Berkeley Lab’s Computational Research Division (CRD) have been awarded massive allocations on the nation’s most powerful supercomputer to advance innovative research in improving the combustion of hydrogen fuels, increasing the efficiency of nanoscale solar cells, and modeling the ignition of Type Ia supernovae. The awards were announced Tuesday, Nov. 30, by Energy Secretary Steven Chu as part of DOE’s Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program. Read more.
HPC Source Features Two Articles about NERSC
The last two issues of HPC Source, a supplement to Scientific Computing magazine, include two articles about NERSC: “Sensing the Future of Greener Data Centers” describes NERSC’s system of wireless sensors and software that helps staff make efficiency improvements and alert operators to environmental issues before they become critical. “Dirac Testbed Reveals How Applications Are Written” describes how NERSC’s Dirac testbed allows users to explore the applicability of graphics processing units to scientific codes.
Supercomputing from Idea to Reality
What’s it like to design and build the biggest computers in the world—system designs that depend on technology that often hasn’t even been invented yet, much less tested? In this hour-long webcast, recorded live at the SC10 conference, Berkeley Lab’s Jonathan Carter and The Register discuss how these systems go from a gleam in an architect’s eye all the way to the TOP500 list. Representatives from Cray, AMD, and Gabriel Consulting also participate. Consider it a crash course in supercomputer procurement from both buyer’s and vendors’ viewpoints.
Slideshow: Berkeley Lab Rolls into New Orleans for SC10
In November 2010, more than 90 staff members from Berkeley Lab Computing Sciences flocked to New Orleans, Louisiana, to participate in tutorials; join panel discussions; lead sessions; give live demonstrations and talks; hold round table discussions; help build SCinet, the world’s fastest network; present posters; and accept prestigious awards. Relive the good times with this slideshow of Berkeley Lab at SC10.
Workshop on Designing and Delivering a Dynamic Research Talk
Have you ever sat through a seminar or group meeting and thought, “This talk is boring. When is it going to end?” Any presentation of your science — from a group meeting to a job talk — offers a key opportunity for you to establish your reputation. In this 90-minute seminar, you’ll learn how to apply professional delivery techniques, effective organizational strategies, and improved slide design to present science with impact. Actual examples from science presentations will be used to demonstrate how to present research — from data slide to scientific story — in a clear and professional manner, leaving your audience wanting to hear more.
The next presentation of this workshop will be 1:00–2:30 pm Thursday, December 9, in the Building 50 auditorium. Go here for more information and registration.
This Week’s Computing Sciences Seminars
Statistical Classification and Level Set Methods in Medical Image Analysis
Monday, December 6, 9:30–10:30 am, 50F-1647
Jenny Folkesson, University of California, San Francisco
With the very large amounts of data produced by medical imaging techniques, there is a growing need for automated analysis. This talk will present examples of automated methods for segmentation and quantitative evaluation of musculoskeletal images using statistical classification, fuzzy clustering, and level set methods. In particular, a method for combining statistical classification and level set methods will be presented with the application of cardiac segmentation.
LAPACK Seminar: Solving Kohn-Sham Nonlinear Eigenproblems
Wednesday, December 8, 11:10 am–12 pm, 380 Soda Hall, UC Berkeley
Yuji Nakatsukasa, UC Davis
There has been considerable progress in recent years in understanding algorithms with inner-outer loop structure for solving extremely large linear eigenvalue problems. To solve nonlinear eigenvalue problems such as those arising from Kohn-Sham electronic structure calculations, a SCF (self-consistent field) eigensolver has a structure of three or more loops. Convergence analysis of such an SCF-eigensolver remains a largely open problem. In this talk, we focus on the convergence analysis of the innermost loop of the SCF-eigensolver. The innermost loop involves solving linear systems of equations and typically consumes most of the total runtime. We propose a scheme to detect the stagnation of the innermost iteration and apply it to improve the computational efficiency of the SCF-eigensolver.
Link of the Week: Putting the Discipline in Interdisciplinary
Many of the most powerful scientific advances are propelled by creative ideas that cross disciplinary boundaries. Unfortunately, initiating interdisciplinary conversations in the context of academia can be a costly undertaking, where disciplines are separated by entrenched physical and social structures. Brainstorming, one of the most common techniques for generating ideas from a diverse group, has been shown by empirical research to be largely ineffective.
Inspired by Malcolm Gladwell’s book Blink, Jeffrey Grossman (a materials science and engineering professor at MIT, formerly at UC Berkeley) and collaborators at UC Berkeley developed and tested a new method called speedstorming, a pair-wise idea generation process that works similarly to the round-robin speed-dating technique. Speedstorming combines an explicit purpose, time limits, and one-on-one encounters to create a setting where boundary-spanning opportunities can be recognized, ideas can be generated at a deep level of interdisciplinary specialty, and potential collaborators can be quickly assessed. By directly comparing speedstorming to brainstorming, they have shown that ideas from speedstorming are more technically specialized and that speedstorming participants are better able to assess the collaborative potential of others. Read more.
About Computing Sciences at Berkeley Lab
The Lawrence Berkeley National Laboratory (Berkeley Lab) Computing Sciences organization provides the computing and networking resources and expertise critical to advancing the Department of Energy's research missions: developing new energy sources, improving energy efficiency, developing new materials and increasing our understanding of ourselves, our world and our universe.
ESnet, the Energy Sciences Network, provides the high-bandwidth, reliable connections that link scientists at 40 DOE research sites to each other and to experimental facilities and supercomputing centers around the country. The National Energy Research Scientific Computing Center (NERSC) powers the discoveries of 6,000 scientists at national laboratories and universities, including those at Berkeley Lab's Computational Research Division (CRD). CRD conducts research and development in mathematical modeling and simulation, algorithm design, data storage, management and analysis, computer system architecture and high-performance software implementation. NERSC and ESnet are DOE Office of Science User Facilities.
Lawrence Berkeley National Laboratory addresses the world's most urgent scientific challenges by advancing sustainable energy, protecting human health, creating new materials, and revealing the origin and fate of the universe. Founded in 1931, Berkeley Lab's scientific expertise has been recognized with 13 Nobel prizes. The University of California manages Berkeley Lab for the DOE’s Office of Science.
DOE’s Office of Science is the single largest supporter of basic research in the physical sciences in the United States, and is working to address some of the most pressing challenges of our time. For more information, please visit science.energy.gov.