InTheLoop | 02.25.2013
The Weekly Newsletter of Berkeley Lab Computing Sciences
February 25, 2013
Video: ESnet Staff Describe How They Contributed to the World’s First SDN Transport
Software Defined Networking (SDN) is an emerging field that makes it easier for software applications to automatically configure and control the various layers of the network. In November 2012, ESnet joined Infinera to demo a prototype SDN Open Transport Switch (OTS), the first capable of dynamically controlling bandwidth services at the optical layer. In this video, ESnet’s Chief Technologist Inder Monga and Network Engineer Chin Guok talk about how they worked with Infinera to reach this milestone. (Monga appears around the one-minute mark).
CS Staff to Share Strategies, Expertise at 100G and Beyond Workshop
ESnet, along with Calit2 (the California Institute for Telecommunications and Information Technology) and CENIC (the Corporation for Education Network Initiatives in California), are convening the 100G and Beyond Workshop, a one-day invited workshop on Tuesday, February 26, at Calit2 on the UC San Diego campus.
The workshop will examine the ways in which 100-gigbabit-per-second networking will impact applications in many areas, including science, health care, and media arts applications; campus and lab strategies; smart manufacturing; network innovation; and regional, national, and international testbeds. ESnet rolled out its 100 Gbps network last November — the first national 100 Gbps science network in the world.
ESnet and NERSC staff will share their expertise in a number of workshop sessions.
- Brent Draney of NERSC will chair and participate in a panel discussion on Campus and Lab Strategies.
- ESnet Chief Technologist Inder Monga will chair the Network Innovations panel, as well as join a discussion on “SDN: Hype vs. Hope.”
- Zarija Lukic of CRD’s Computational Cosmology Center will also join the SDN discussion, and present on “Future Coupling of Major Instruments: NERSC and SDSC.”
- ESnet’s Brian Tierney will describe “Deploying and Testing Networking Innovations on a Wide Scale: ESnet” as part of a panel on Scientific Workflows and Testbeds.
- ESnet Director Greg Bell will join a discussion on next steps in the closing session, “From Campus Innovation to California/National Testbeds.”
CS Staff Contribute to SIAM Conference on Computational Science and Engineering
The annual SIAM Conference on Computational Science and Engineering is being held this week, February 25–March 1, in Boston, MA. The SIAM CS&E conference seeks to enable in-depth technical discussions on a wide variety of major computational efforts on large problems in science and engineering, foster the interdisciplinary culture required to meet these large-scale challenges, and promote the training of the next generation of computational scientists.
Contributions to the conference from Berkeley Lab Computing Sciences researchers are listed below. (Co-authors from other institutions are omitted.)
- Ann S. Almgren, John B. Bell, and Michael Lijewski: BoxLib: Overview and Applications
- John B. Bell and Kaushik Balakrishnan (co-authors): Modeling of Thermal Fluctuations in Multicomponent Reacting Systems
- Aydin Buluc (co-organizer): Minimizing Communication in Scientific Computing
- Michael Driscoll, Evangelos Georganas, Penporn Koanantakool, and Katherine Yelick (co-authors): A Communication Optimal N-Body Algorithm for Long-Range Direct Interactions
- Phillip Collela and Peter Mccorquodale (co-authors): Gyrokinetic Edge Plasma Simulation Using Continuum Methods
- James W. Demmel (co-author): Avoiding Communication in Parallel Bidiagonalization of Band Matrices
- James W. Demmel (co-author): Shape-Morphing in LU Factorizations
- James W. Demmel (co-author): Lower Bounds on Algorithm Energy Consumption: Current Work and Future Directions
- James W. Demmel and Shoaib Kamil (co-authors): Beating MKL and Scalapack at Rectangular Matrix Multiplication Using the BFS/DFS Approach
- Anthony Leroy Drummond (co-author): Auto-Tuning and Smart-Tuning Approaches for Efficient Krylov Solvers on Petascale Architectures
- Matthew Emmett (co-organizer): Space-Time Parallel Methods: Algorithms, Implementation and Applications
- Matthew Emmett (co-author): Implications of the Choice of SDC Nodes in the Multilevel PFASST Algorithm
- John R. Gilbert, Aydin Buluc, Shoaib Kamil, Adam Lugowski, Leonid Oliker, and Samuel Williams (co-authors): High-Performance Filtered Queries in Attributed Semantic Graphs
- August Johansson: A High Order Discontinuous Galerkin Nitsche Method for Elliptic Problems with Fictitious Boundary
- Hans Johansen, Phillip Colella, and Peter Mccorquodale (co-authors): Adaptive Fourth-Order Cubed Sphere Discretization for Non-Hydrostatic Atmosphere Simulations
- Alice Koniges: Programming Model Exploration and Efficiency Modeling using Mini and Proxy Applications
- Xiaoye Sherry Li: Recent Advances in Scalable Sparse Factorization Methods
- Lin Lin and Chao Yang: Elliptic Preconditioner for Accelerating the Self Consistent Field Iteration of Kohn-Sham Density Functional Theory
- Osni A. Marques and Leroy A. Drummond (co-organizers): Auto-Tuning Technologies for Tools and Development Environment in Extreme-Scale Scientific Computing
- Osni A. Marques (co-author): Assessing Library Performance with TAU
- Daniel Martin and Esmond Ng (co-authors): Resolving Grounding Line Dynamics Using the BISICLES Adaptive Mesh Refinement Model
- Matthias Morzfeld and Alexander J. Chorin: Implicit Particle Methods for Data Assimilation
- Andy Nonaka and John B. Bell (co-authors): Low Mach Number Fluctuating Hydrodynamics of Diffusively Mixing Fluids
- Per-Olof Persson (chair): Numerical Methods for PDEs
- Per-Olof Persson: Shock Capturing for High-Order Discontinuous Galerkin Simulation of Transient Flow Problems
- Per-Olof Persson and Bradley Froehle: A High-Order Implicit-Explicit Discontinuous Galerkin Scheme for Fluid-Structure Interaction
- Chris Rycroft: Mechanical Simulation of Mammalian Acini
- David Trebotich: Pore Scale Reactive Transport Modeling using Adaptive, Finite Volume Methods with a Look toward Upscaling
- Didem Unat (co-author): Mint: A User-Friendly C-to-CUDA Code Translator
- Ethan Van Andel, Ann S. Almgren, John B. Bell, and Michael Lijewski: Region-Based AMR: A New AMR Paradigm in BoxLib
- Eugene Vecharynski (co-author): Absolute Value Preconditioning for Symmetric Indefinite Linear Systems
- Eugene Vecharynski (co-author): Updating Singular Subspaces for Latent Semantic Indexing
- Gunther H. Weber and Dmitriy Morozov: Geometric Comparisons in Porous Media Simulation
- Jon Wilkening: Stability of Interacting Solitary Water Waves, Standing Waves, and Breathers
- Chao Yang (session organizer): Large-scale Eigensolvers for Many-/Multi-Core Systems
- Chao Yang: Acceleration Techniques for Electronic Structure Calculation
- Chao Yang (co-author): Computing Eigenspace by a Penalty Approach
- Chao Yang and Esmond Ng (co-authors): An Efficient and Scalable Lanczos-based Eigensolver for Multicore Systems
- Chao Yang, Hasan Metin Aktulga, and Lin Lin(co-authors): Computing a Large Number of Eigenpairs on Multi-/Many-Core Systems
MSRI and Monday Night PlayGround Present “Fearful Symmetry” Tonight
Six short new plays inspired by the topic “Fearful Symmetry” will be presented at 8:00 pm tonight, February 25, at the Berkeley Repertory Theatre, 2025 Addison Street in downtown Berkeley, sponsored by the Mathematical Sciences Research Institute (MSRI) and Monday Night PlayGround. The PlayGround (now in its 19th season) is a monthly presentation of six short plays written on a common theme.
On February 13, two mathematicians at MSRI, Dave Benson and Robert Bryant, met with the PlayGround pool of playwrights to talk about symmetry in mathematics and music, how mathematicians think about them, and the often surprising ways in which symmetry appears in our world and in our culture. The playwrights then had five days to write a short play inspired by the topic, and the best six plays will be performed tonight.
A pre-show discussion, which starts at 7:10 pm, will feature a panel of mathematicians and playwrights. More information, including where and how to buy tickets, can be found here.
Climate Disruption: What Math and Science Have to Say
This year, 2013, has been declared the Year of the Mathematics of Planet Earth (MPE 2013) by the International Mathematical Union. The Mathematical Sciences Research Institute and the American Institute of Mathematics are hosting “Climate Disruption: What Math and Science Have to Say” at 7:30 pm on Monday, March 4, at the Palace of Fine Arts Theater in San Francisco, as part of an international public lecture series sponsored by the Simons Foundation.
The speaker will be Emily Shuckburgh, professor of mathematics at Cambridge University and leader of the British Antarctic Survey’s Open Oceans Research Group, which is focused on understanding the role of the polar oceans in the global climate system. Her personal research concerns investigating the dynamics of the atmosphere, oceans and climate using theoretical approaches, observational studies and numerical modeling.
This Week’s Computing Sciences Seminars
Exploit Mitigation: From Detection to Obstruction — Challenges and Vision: Or Why Trying to Identify Unknown Attacks Is As Inefficient As It Sounds
Tuesday, February 26, 1:00–2:00 pm, 430 Soda Hall (Wozniak Lounge), UC Berkeley
Gal Badishi, Chief Scientist, Cyvera, and Uri Alter, CEO and Co-Founder, Cyvera
Recent attacks on high-value targets demonstrate how state-of-the-art defenses fail to protect against APTs (Advanced Persistent Threats). These victims spare no expense and appropriately deploy cutting-edge defenses, such as firewalls, intrusion detection and prevention systems, and anti-virus scanners, as well as novel approaches for detecting zero-day exploits - yet these are ineffective at thwarting determined attackers.
This talk will examine the unsatisfying state of attack-prevention solutions, and we will demonstrate the ease of circumventing the majority of defenses.
We move on to present a fresh security paradigm: extensive obstruction of attacks, rather than an attempt to identify and detect malicious behaviors and attack-related actions, often after the fact. In combining methods such as traps in heap memory and DLL protection, with enhancements to solutions as Data Execution Prevention (DEP) and Address Space Layout Randomization (ASLR), we achieve nearly perfect exploit-prevention rates, even for zero-day exploits.
Further, we’ll discuss the challenges in transforming these mitigation techniques into a commercial-grade product with security modules that can be applied generically to every process.
Measuring the I/O Load on the Hopper Lustre File System
Wednesday, February 27, 12:00–1:00 pm, OSF 943-238
Andrew Uselton, NERSC
The Hopper scratch file systems have a peak data rate of 35 GB/s, and the Lustre Monitoring Tool (LMT) shows that this peak is achieved in practice. LMT also shows that most of the time the file systems are moving data at a much lower rate than that. It is well known that poorly organized I/O will not achieve optimum performance. The file system can be heavily loaded without the load being apparent in the LMT view. A new extension to LMT now allows us to capture the I/O characteristics and thereby show a truer view of the load on the file system. This talk presents the findings from reviewing four months of data from the newly augmented LMT utility.
Gaussian Smoothing for Nonconvex Optimization with Applications to Computer Vision
Friday, March 1, 12:00–1:00 pm, 510 Soda Hall, UC Berkeley
Hossein Mobahi, Massachusetts Institute of Technology
The Gaussian function is the center of several relaxation methods for solving difficult optimization problems. When seen as a filter, its smoothing property can be used for eliminating spurious local minima. The idea is to start from a highly smoothed version of the objective that is hopefully easier to minimize. Then, the path of that solution is followed as the function is gradually deformed back to its original shape.
There are fundamental questions around such continuation method that are not well understood. For example, what functions eventually become convex after enough smoothing? For such a function, does its asymptotic minimizer have a simple closed form? I present, application independent and easy to check conditions, related to these properties.
In addition to these topics, there are other issues that matter from practical viewpoint. For example, if the optimization space is high dimensional, then smoothing becomes expensive due to the curse of dimensionality in numerical integration. However, I show that sometimes the high-dimensional convolution can be equivalently expressed by a low-dimensional integral operator. I present such operators for the tasks of image alignment and 3D point cloud registration (although the underlying concept may be extendable to other optimization tasks even beyond computer vision). Finally, I will briefly discuss my ongoing research on using this optimization method for learning visual object representation.
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.