InTheLoop | 10.17.2011
October 7, 2011
ESnet, Internet2 Complete 100G Network Deployment
Two of the nation’s leading research networks—the U.S. Department of Energy’s (DOE) Energy Sciences Network (ESnet) and Internet2—last week announced that they have completed the world’s first transcontinental deployment of 100 Gigabit per second (Gbps) using coherent technology. Built on Ciena’s 6500 Packet-Optical Platform, the new 8.8 Terabit per second network is equipped with 100 Gbps optical backbone connections. Those connections are now operational between New York, Washington, D.C., Cleveland, Chicago, Kansas City, Denver, Salt Lake City, and Sunnyvale, spanning a distance nearly 4000 miles. Read more.
Computational Research Division Announces Reorganization
In an effort to balance the size and research areas of the departments within the Berkeley Lab’s Computational Research Division (CRD), Division Director David Brown announced a reorganization last week.
John Bell will lead the new Mathematics and Computational Science Department, comprising the Center for Computational Sciences and Engineering, the Mathematics Group, and the Computational Cosmology Center. Esmond Ng will lead the new Applied Mathematics and Scientific Computing Department, comprising the Applied Numerical Algorithms Group, the Scientific Computing Group, and the Complex Systems Group. John Shalf will lead the new Computer and Data Sciences Department, comprising the Visualization Group, the Scientific Data Management Group, the Future Technologies Group, and the Advanced Technologies Group.
The Advanced Computing for Science Department will continue to be led by Deb Agarwal, and Victor Markowitz will continue his leadership of the Biological Data Management and Technology Center. Read more.
NERSC to Deploy Redesigned NIM Interface
On Wednesday afternoon, October 19, NERSC will be releasing a redesigned interface for the NERSC Information Management (NIM) online database. The new release will feature a more user-friendly interface without frame borders, and including improvements such as navigation moved to the top of the page, and repo balances and system status on the first page after login. Performance improvements have also been made to reduce page load times, especially for the most common NIM functions. Clayton Bagwell is NIM Development Team Leader. Read more.
Experimental Mathematics: Computing Power Leads to Insights
In an article called “Exploratory Experimentation and Computation,” appearing in the November 2011 issue of the Notices of the American Mathematical Society, David Bailey, head of CRD’s Complex Systems Group, and his colleague Jonathan Borwein of the University of Newcastle, Australia, describe how modern computer technology has vastly expanded our ability to discover new mathematical results. In addition to discussing state-of-the-art uses of computers in mathematics, the article also touches on the need to refashion mathematics education to give students the tools of experimental mathematics. Read more.
CRD Researchers Contribute to DOE Applied Math Program Meeting
The 2011 DOE Applied Mathematics Program Meeting is being held this week, October 17–19, at the Hyatt Regency Reston in Washington, DC. The 2.5-day meeting will consist of a combination of plenary talks, breakout sessions, and poster sessions to highlight recent research in selected theme areas. Presentations by CRD researchers include:
- John Bell: Applied Math Workshop Update (with Mihai Anitescu)
- John Bell: Finite volume methods for fluctuating hydrodynamics
- Phillip Colella: Adaptive High-Order Finite-Volume Methods for Partial Differential Equations
- Jim Demmel: Minimizing Communication in Linear Algebra (plenary session)
- Ming Gu: Rapid and Reliable Randomized Algorithms
- Xiaoye (Sherri) Li: Towards an Optimal Parallel Approximate Sparse Factorization Algorithm Using Hierarchically Semi-Separable Structures
- Andy Nonaka: A Deferred Correction Strategy for Advection-Diffusion-Reaction Systems with Complex Chemistry
- Christopher Rycroft: Application of the Voronoi tessellation for high-throughput analysis of crystalline porous materials
- James Sethian: A New Framework for Tracking Multiphase Physics: Foams, Membranes and Cells
- Sean Whalen: Classifying Parallel Computation with Network Theory & Machine Learning
- Jon Wilkening: Breakdown of self-similarity at the crests of large amplitude tanding water waves, and other applications of overdetermined shooting methods
Poster presentations include:
- Alexandre Chorin: Implicit sampling with applications to filtering and data assimilation
- Orianna DeMasi: A Numerical Study of Three Ensemble Methods
- Steven Hofmeyr: Modeling Interventions in Complex Networks
- Per-Olof Persson: Sparse Line-Based Discontinuous Galerkin Discretizations and Efficient Time-Integration
- David Trebotich: An Adaptive Embedded Boundary Method for Pore Scale Reactive Transport
NERSC Staff Contribute to HPSS User Forum
The HPSS User Forum (HUF 2011) is being held this week, October 18–21, at Indiana University in Bloomington. HUF 2011 is the annual gathering for the HPSS community, bringing together new and existing HPSS users from around the globe to discuss best practices, new implementations, and future directions and releases. Mike Welcome of NERSC will give a presentation on “MPS—Advice, Changes and Future Direction,” and Nick Balthaser will give the NERSC Site Report.
Matlab Technical Sessions Today on the Hill
Mathworks application engineer Isaac Noh is presenting two Matlab technical sessions today in the Bldg. 50 Auditorium. Registration is requested but not mandatory. Here is the agenda:
9:45–10:00 am: Registration and sign-in. Walk-ins are welcome.
10:00 am–12:00 pm: Session 1: Technical Computing with MATLAB: Test and Measurement
Attend this free seminar to find out how you can use MATLAB and its add-on products to develop algorithms, visualize and analyze data, and perform numeric computation. MathWorks engineers will provide an overview of MATLAB through live demonstrations, examples, and user testimonials, showing how you can use MATLAB and related toolboxes to:
- Access data from many sources (files, other software, hardware, etc.)
- Use interactive tools for iterative exploration, design, and problem solving
- Automate and capture your work in easy-to-write scripts and programs
- Share your results with others by automatically creating reports
- Build and deploy GUI-based applications
Many applications require access to live or real-world data from external devices. You will also see how you can:
- Connect to instruments and data acquisition cards from MATLAB
- Acquire live signals, images, and video inside MATLAB
- Use MATLAB to build and deploy a test application
MATLAB provides a flexible environment for teaching and research in a wide range of applications, including signal processing and communications, image processing, math and optimization, statistics and data analysis, control systems, hardware data acquisition, computational finance, and computational biology.
This seminar is appropriate for attendees with beginner to expert MATLAB experience.
12:00–12:30 pm: A light lunch will be served.
12:30–2:30 pm: Session 2: Handling Large Data Sets
This seminar will describe strategies for handling large amounts of data in MATLAB and avoiding out-of-memory errors. It will provide you with an understanding of the causes of memory limitations in MATLAB and a set of techniques to increase the available memory in MATLAB. It will also show techniques for minimizing memory usage in MATLAB while accessing, storing, processing, and plotting data.
Additionally, accessing memory across multiple systems using parallel computing will be explored.
- Understanding memory and its constraints
- Minimizing your memory footprint in MATLAB
- Accessing memory across multiple systems
This Week’s Computing Sciences Seminars
Statistics and Computation in the Age of Massive Data
Monday, October 17, 4:00–5:00 pm, 430 Soda Hall (Wozniak Lounge), UC Berkeley
Michael Jordan, UC Berkeley EECS
There are many issues remaining to be addressed, or even formulated, at the Interface of statistics and computation. One way to capture the current state of affairs is the following: If we view data as a resource, how can it be that in many practical problems of interest we find ourselves embarrassed by being given too much data? Our inferential procedures use polynomial amounts of time and space but that doesn't suffice; we need to be able to guarantee that on a fixed computational budget the statistical risk decreases as the number of data points grows (without bound). A general theory not yet being available, in this talk I present three vignettes that describe various lines of attack on the problem: one involving the bootstrap, another involving matrix completion algorithms, and the third involving phylogenetic analysis in the regime of large numbers of taxa. All three vignettes involve divide-and-conquer strategies, with the third vignette being particularly interesting in this regard (divide-and-conquer arises from Poisson thinning). [Joint work with Alex Bouchard-Cote, Ariel Kleiner, Lester Mackey, Purna Sarkar and Ameet Talwalkar.]
Managing Your Personal NERSC Staff Web Page and Using the Web Publications Database
Tuesday, October 18, 12:00–1:30 pm, OSF Room 238
Margie Wylie, LBNL Computing Sciences Communications Group
Every NERSC staff member has a personal web page on the new NERSC web site. Margie will explain how to modify your page, add publications and presentations to the web database (which will automatically display on your staff page), and share web authoring best practices. She will also be available for general Q&A about authoring web pages on our site.
Past, Present and Future of LibGeoDecomp, an Auto-Parallelizing Stencil Code Library
Wednesday, October 19, 2:00–3:00 pm, 50F-1647
Andreas Schäfer, Friedrich-Alexander-Universität, Erlangen-Nürnberg, Germany
Stencil codes represent the vast majority of computer simulations, with applications ranging from astrophysics to crystal growth. Given their tightly coupled nature and the increasing variety of today’s parallel computers, their parallelization remains challenging. In this talk I will give an overview of how LibGeoDecomp (Library for Geometric Decomposition codes), a C++ class template library, can help users to parallelize their codes without having to deal with most technical hurdles, e.g., harnessing MPI and CUDA. In the second part I will outline how the library could/should be extended to make it useful for a wider audience.
Novel All-Spin Devices and Architectures for Low Power Computing
Friday, October 21, 1:00–2:00 pm, 521 Cory Hall (Hogan Room), UC Berkeley
Behtash Behin-Aein, GLOBALFOUNDRIES
Electronic devices have traditionally been based on controlling the flow of charge. However, electrons carry both charge and spin, the latter being responsible for magnetic phenomena. In the last ten years there have been significant advances in our ability to control the spin current in electronic devices and their interactions with nanomagnets. All-spin logic (ASL) represents a new approach to information processing where the roles of charges and capacitors in CMOS are played by spins and magnets. ASL is defined by: Information: Magnetization of magnet, Communication: Spin current, Energy: Power supply. In this talk we will (1) present our recent progress on the design of input-output isolation and intrinsic directivity of ASL devices, (2) show the simulation based operation of a NAND gate with Fan-out of 2 that propagates information without clocking. This is based on the ASL benchmarked simulator for multi-magnet networks coupled by spin transport channels, and (3) describe the natural mapping of such ASL networks into neuromorphic circuits suitable for hybrid analog/digital information processing.
Link of the Week: Another Computer Legend Passes
Dennis M. Ritchie, who helped shape the modern digital era by creating software tools that power things as diverse as search engines like Google and smart phones, died on October 12 at his home in Berkeley Heights, N.J. He was 70.
As principal designer of the C programming language and co-developer of the UNIX operating system, Ritchie helped create the foundation for much of modern computing. “The tools that Dennis built — and their direct descendants — run pretty much everything today,” said Brian Kernighan, a computer scientist at Princeton University who worked with Mr. Ritchie at Bell Labs.
For all of Ritchie's influence on the industry, he never really became a pop icon in the manner of Steve Jobs or Bill Gates. Unlike Jobs and Gates, who led their respective companies to fame and fortune, Ritchie was the archetypal computer scientist — the guy that came up with all the great ideas, upon which others built great empires. To programmers of his time, though, he was a hero.
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 7,000-plus 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 Department of Energy 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.