InTheLoop | 10.24.2011
October 24, 2011
ESnet and LGS Deploy Alcatel-Lucent Routers for New 100G Network
LGS Innovations last week (Oct. 19) announced it is teaming with ESnet to build a nationwide 100 Gigabit per second Ethernet (100GE) network that will accommodate ESnet’s anticipated tenfold increase in science data traffic volume over the next four years and beyond. The network will support research efforts that will enable scientists to develop technologies and innovations that will address energy, climate, and other research challenges in the U.S. and abroad. ESnet is building the new 100GE network through its Advanced Networking Initiative (ANI).
Under the agreement, LGS Innovations is providing the 7750 Service Router (SR) from its parent company, Alcatel-Lucent, for ten points of presence across ESnet’s national backbone as well as directly at select laboratory sites. LGS is providing this solution through a competitively awarded subcontract with Synchronized Networking Solutions, LLC. Read more.
perfSONAR pS-Performance Toolkit 3.2.1 Released
Need to troubleshoot some network performance problems? perfSONAR-PS is an open source development effort to create a collection of easy-to-use and easy-to-install perfSONAR network performance monitoring services and tools. Version 3.2.1 of the pS-Performance Toolkit is now available for download. This update contains new throughput graphs, new delay/loss graphs, scheduled traceroute tests, and numerous bug fixes. Release notes are available to denote all changes since the 3.2 release. Please visit the pS Performance Toolkit page for more information.
Leading ESnet to the Next Level of Innovation
Scientific Computing’s HPC Innovator Series recently published an interview with ESnet Department Head Steve Cotter titled “Leading ESnet to the Next Level of Innovation.” Among other topics, Cotter discusses the need for more energy-efficient networking.
“In aggregate, a lot of energy is being used by networks, but no one is really looking at that,” Cotter said. “This is a field in which we can have a real impact and kick start a new way of thinking…. Although we are a Department of Energy facility, I’d like us to become the Department of Less Energy.” Read more.
Berkeley Lab Well Represented at 2011 ASCR/BES Data Workshop
The U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences and Office of Advanced Scientific Computing Research are holding the 2011 ASCR/BES Data Workshop today and tomorrow, October 24–25, in Bethesda, MD. The goal of this workshop, with the theme “Data and Communications in Basic Energy Sciences: Creating a Pathway for Scientific Discovery,” is to identify current and anticipated issues in the acquisition, analysis, communication and storage of experimental data that could impact the progress of scientific discovery.
NERSC Analytics Group Lead Peter Nugent is co-chairing the workshop with Michael Simonson of Oak Ridge National Lab. Associate Lab Director Kathy Yelick and ESnet’s Bill Johnston will participate in the Workflow Management panel. Chao Yang of CRD will co-chair the Theory and Algorithms panel, with the participation of CRD Division Director David Brown, Stefano Marchesini of the Advanced Light Source (ALS), and Rob Ryne of the Accelerator and Fusion Research Division.
Participants in the Visualization and Analysis panel will include Wes Bethel and Craig Tull of CRD, Alex Hexemer of the ALS, Frank Ogletree of the Materials Science Division, and Jim Ciston of the National Center for Electron Microscopy. Eli Dart of ESnet will co-chair the Data Processing and Management panel; participants include Shane Canon of NERSC, Peter Denes of the Lab Directorate, Michael Banda of the ALS, Arie Shoshani of CRD, and Lauren Rotman of ESnet.
CRD Staff Share Expertise at VisWeek
VisWeek 2011, which is taking place October 23–28 in Providence, RI, is the premier forum for visualization advances for academia, government, and industry, bringing together researchers and practitioners with a shared interest in tools, techniques, technology and theory. The week is organized around three separate conferences: IEEE Visualization, IEEE Information Visualization, and IEEE Visual Analytics Science and Technology.
David Camp of CRD is giving two talks that were co-authored with Hank Childs along with researchers from other institutions: “Streamline Integration Using MPI-Hybrid Parallelism on a Large Multicore Architecture,” and “Evaluating the Beneﬁts of an Extended Memory Hierarchy for Parallel Streamline Algorithms.” Gunther Weber will chair a session on “Statistics, Geometry and Signal Processing.” And Kesheng John Wu will present a paper on “Parallel In Situ Indexing for Data-Intensive Computing,” co-authored with Jinoh Kim and Arie Shoshani of CRD and collaborators from other institutions.
RCE Interviews R. K. Owen on Environment Modules
The Modules software environment is a significant but very much under-the-hood part of how NERSC provides services to so many users, and it has become a fixture at computing centers around the world. R. K. Owen of NERSC took over development and administration of the open-source project many years ago, keeping the project current and vibrant, adding features, and making it more robust.
Modules is not fancy software. It does something very basic but which is also crucial to effectively managing a user base of many thousands of scientists: it allows many users with different preferences to see the same set of software the way they want to see it, and to not see things that are irrelevant to them. That may sound humdrum, but Modules plays a big role in making supercomputers easier to use.
Owen was recently interviewed about his role in developing this software by Jeff Squyres of Cisco Systems and OpenMPI fame, and Brock Palen of the University of Michigan Center for Advanced Computing, who together host the RCE (Research Computing and Engineering) podcast.
Webinar to Offer Tips on Recruiting Women in Engineering
The Women in Engineering Program Advocates Network (WEPAN) and Mentornet.net are presenting a free webinar on “Mentoring Millennials: Evolving practices for guiding a new generation of women engineers to career success” from 10:00 to 11:00 am on Tuesday, October 25.
Millennial students differ from their predecessors in important ways that can assist us in recruiting and retaining more women in engineering. For example, millennial students, male and female, are genuinely interested in making a difference in the world. Using initiatives such as the NAE's Grand Challenges for Engineering, we can clearly link engineering with that desire to make a difference.
This Week’s Computing Sciences Seminars
HMPP: A Directive Based Approach to Address GPUs
Tuesday, Oct. 25, 10:00–11:00 am, 50B-2222
Denis Gerrer, CAPS Enterprise
This talk will give you an introduction on how to use HMPP to address GPU (NVIDIA and/or ATI) with your legacy application. HMPP is a source-to-source compiler that generates CUDA and/or OpenCL code from existing C or Fortran code that offers a high level abstraction for hybrid programming.
HMPP compiler integrates powerful data-parallel backends for NVIDIA CUDA and OpenCL that drastically reduce development time. The HMPP runtime ensures application deployment on multi-GPU systems. Software assets are kept independent from both hardware platforms and commercial software. While preserving portability and hardware interoperability, HMPP increases application performance and development productivity.
This talk will expose you to the main features of HMPP, making you able to write your own hybrid applications and use HMPP in an incremental way to move your application to GPUs. In this talk you will have an overview of the two sets of directives to port your application to GPU: first set to generate a kernel and move data to the GPU, and second set to optimize the GPU kernel.
Bridging Photonics and Computing
Wednesday, Oct. 26, 12:00-1:00 pm, 310 Sutardja Dai Hall (Banatao Auditorium), UC Berkeley
Live broadcast at mms://media.citris.berkeley.edu/webcast
Mario Paniccia, Director of Photonics Technology Lab, Intel
The silicon chip has been the mainstay of the electronics industry for the last 40 years and has revolutionized the way the world operates. Today a silicon chip the size of a fingernail contains over one billion transistors and has the computing power that only a decade ago would take up an entire room of servers. Recently silicon photonics has attracted a great deal of attention since it offers an opportunity for low cost opto-electronic solutions for applications ranging from telecommunications down to chip-to-chip interconnects as well as possible applications in new emerging areas such as optical sensing and or bio-medical applications.
Recent advances and research breakthroughs in silicon photonic device performance over the last few years have shown that silicon can be considered as a material onto which one can build future optical devices. While significant efforts are needed to improve device performance and to “commercialize” these technologies, progress is moving at a rapid rate. If successful, silicon may similarly come to impact optical communications as it has impacted the electronics industry.
This presentation will provide an overview of silicon photonics research at Intel Corporation, describe some of the recent advances including the recently announced demonstration of an integrated silicon photonics optical link operating at 50Gbps and the scalability of this technology to >1Tbps. In addition the presentation will provide an overview and discuss the potential applications and future opportunities for enabling “photonics” in and around the PC and server platform.
State Estimation in Large-Scale Open Channel Networks Using Sequential Monte Carlo Methods: Optimal Sequential Importance Resampling and Implicit Particle Filters
Wednesday, Oct. 26, 4:10-5:00 pm, 939 Evans Hall, UC Berkeley
Mohammad Rafiee, UC Berkeley
We investigate the performance of sequential Monte Carlo estimation methods for estimation of flow state in large-scale open channel networks. After constructing a state space model of the flow based on the Saint-Venant equations, we implement the optimal sequential importance resampling (SIR) filter to perform state estimation in a case in which measurements are available every time step. Considering a case in which measurements become available intermittently, a random map implementation of the implicit particle filter is applied to estimate the state trajectory in the interval between the measurements. Finally, some heuristics are proposed which are shown to improve the estimation results and lower the computational cost. In the first heuristics, considering the case in which measurements are available every time step, we apply the implicit particle filter over time intervals of a desired size while incorporating all the available measurements over the corresponding time interval. As a second heuristic, we apply an approximate maximum a posteriori (MAP) method which does not require sampling. It is seen, through implementation, that the MAP method provides more accurate results in the specific case of our application while having a smaller computational cost. All estimation methods are tested on a network of 19 subchannels and one reservoir, Clifton Court Forebay, in Sacramento-San Joaquin Delta in California, and numerical results are presented.
Transition from Planar MOSFETs to FinFETs and Its Impact on Design and Variability
Friday, Oct. 28, 1:00–2:00 pm, 521 Cory Hall (Hogan Room), UC Berkeley
Victor Moroz, Synopsys, Inc.
Analysis of the transistor evolution happening in the industry over the last several technology nodes reveals that despite the on-schedule chip area scaling, there has been a crisis in transistor scaling. In leading edge technology, the critical transistor size did not change all the way from 90nm to 20nm nodes. The underlying physical mechanisms explain why it is happening and point to the upcoming changes in transistor architecture that will enable transistor shrinking to resume. The scaling crisis is responsible for the slower than anticipated variability increase and for the stress engineering taking over the driver's seat in the performance race. The change in transistor architecture at 20nm and 15nm nodes will change the balance of major variability mechanisms and some of the lithography requirements and design rules. It will also open the door for introduction of non-silicon transistors built on top of a silicon wafer. Comparative analysis of a planar bulk MOSFET with FDSOI MOSFET and a FinFET shows their pros and cons and suggests their likely roles in the future.
High Order Mimetic Differential Operators
Friday, Oct. 28, 2:00–3:00 pm, 50F-1647
Jose Castillo, San Diego State University
We will present advancements on the theory and application of High Order Mimetic Difference Operators. The main goal of this research is to construct local high order difference approximations of differential operators on nonuniform grids that mimic the properties of the continuum operators. Partial differential equations solved with these mimetic difference approximations often automatically satisfy discrete versions of conservation laws and analogies to Stokes’ theorem that are true in the continuum and as a consequence are more likely to produce physically faithful results.
Link of the Week: NSF Supports Family Time
Internal changes to a government agency’s home-and-work policy don’t normally warrant a White House rollout and an accompanying Washington Post op-ed. But when the National Science Foundation unveiled plans to instill comprehensive support for work-family balance throughout the foundation and its grant work, it was a big step toward redressing the gender gap in U.S. science and engineering.
The new policy isn’t geared exclusively toward women, although they will be its most practical beneficiaries. Researchers will be able to extend or delay research grants to have a baby or adopt a child. They will be able to take time off to care for elderly family members without worrying about losing grant money, research projects or (hopefully) career status. The NSF will even provide funding for research technicians to carry on their work while a parent temporarily leaves the lab. 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.