InTheLoop | 05.03.2010
May 3, 2010
NERSC Users Manipulate Graphene with Nanodroplets of Water
Chemists using NERSC’s Cray XT4 supercomputer, Franklin, have found that nanodroplets of water can be used to shape graphene into complex capsules, sandwiches, knots, and rings to serve as the building blocks of nanodevices with unique mechanical, electrical, and optical properties. Read more.
Why Does Matter Dominate Our Universe?
Using NERSC systems, William Detmold of DOE’s Thomas Jefferson National Accelerator Facility and his colleagues achieved the first ever QCD calculations for both a three-body force between hadrons and three-body baryon system. In a subsequent paper, Detmold and Martin Savage also reported another first: a QCD calculation that could help scientists better understand the quark soup that was our universe milliseconds after its birth. Read more.
NERSC, ESnet and JGI Explore Hardware as a Service
NERSC, ESnet, and the Joint Genome Institute (JGI) are showing that a “hardware as a service” model, where computing resources are provisioned as needed, could be quickly deployed to meet the computing demands of the research community. Read more.
CS Hosts DOE Applied Math Program Meeting This Week in Berkeley
Berkeley Lab Computing Sciences is hosting the 2010 DOE Applied Mathematics Program Meeting today through Wednesday, May 3–5, at the Doubletree Hotel Berkeley Marina. Esmond Ng chairs the Organizing Committee, and Chao Yang is a committee member. Eighteen CRD researchers are giving presentations or posters:
- Andy Aspden: “Convective Pattern Formation During Rotation Spin-Up” (poster)
- Alexandre Chorin: “Implicit Particle Filters: Theory and Applications”
- Marc Day: “Transition from Cellular to Distributed Burning in Lean Hydrogen-Air Flames” (poster)
- Jeffrey Donatelli: “Numerical Methods for Incompressible Two-Phase Flow in Complex Geometries” (poster)
- Aleksandar Donev: “On the Accuracy of Finite-Volume Schemes for Fluctuating Hydrodynamics”
- Maciej Haranczyk: “PDE-Based Analysis of Void Space of Porous Materials Used in Energy-Related Applications” (poster)
- Steven Hofmeyr: “ASIM: An Agent-Based Integrated Model for Complex Systems”
- Jakub Kominiarczuk: “Algebraic Monte Carlo without Markov Chains” (poster)
- X. Sherry Li: “On Techniques to Enhance Robustness and Scalability of the Schur Complement Method” (poster)
- Matthias Morzfeld: “Nonlinear Data Assimilation with an Implicit Filter for Stochastic Partial Differential Equations Exhibiting Space-Time Chaos” (poster)
- George Pau: “High Resolution Simulation and Characterization of Flows Encountered in Geologic CO2 Sequestration” (poster)
- Marcus Roper: “Hierarchical Models of Flagellum Dynamics for Device-Design and for Understanding Evolution of Cell-Cell Cooperation” (poster)
- Chris Rycroft: “Voro++: A Software Library for Three-Dimensional Voronoi Cell Calculations” (poster)
- James Sethian: “New Algorithms for Physics across Moving Interfaces: Fluid Breakup, Coating Rollers and Two-Phase Flow”
- Horst Simon: “Exascale Challenges for the Applied Mathematics Community”
- Daniela Ushizima: “Vasculature Segmentation of Ocular Fundus using Fast Marching” (poster)
- Jon Wilkening: “Computation of Time-Periodic Solutions of Fluid Interface Problems” (poster)
- Chao Yang: “Selected Inversion of Sparse Matrices with Application to Electronic Structure Calculations” (poster)
Aydin Buluç Joins CRD As 2010 Alvarez Fellow
As the 2010 Luis W. Alvarez Fellow and a member of the High Performance Computing Research Department, Aydin Buluç will apply his expertise in combinational scientific computing primarily to increasing the reliability of the electrical grid and improving the nation’s ability to respond to energy disruptions. His research interests include parallel computing, combinatorial scientific computing, high performance graph analysis, and sparse matrix computations. Read more.
Shoshani Talks on Scientific Data Management Challenges at Salishan
The Salishan Conference on High-Speed Computing met last week at Salishan Lodge in Gleneden Beach, Oregon. Arie Shoshani, head of the Scientific Data Management Research Group in CRD, gave an invited lecture on “Scientific Data Management Challenges in Extreme Scale Systems.”
Berkeley Water Center Featured at Microsoft’s Silicon Valley TechFair
The Berkeley Water Center, a joint effort between LBNL’s Advanced Computing for Science Department and UC Berkeley with funding from Microsoft Research, will be highlighted at the May 6 Silicon Valley TechFair, held on Microsoft’s Mountain View campus. The work will be highlighted in the keynote address at the TechFair, and will be featured in both a demo and a five-minute video.
The Berkeley Water Center promotes and supports collaborative, water-related research within the Berkeley research community. The LBNL project team has helped developed a data server providing an integrated view of discharge, precipitation, stream temperature and air temperature across California watersheds. This data server is now being used to support researchers working on a variety of questions including impact of frost protection pumping, recovery of endangered fish populations, long-term impact on a watershed of human activity, coastal lagoon dynamics, and modeling of annual watershed water balance. The work has also been used to help develop the statewide recovery plan for California’s wild Coho salmon.
David Bailey to Give Keynote Talk at Canadian Research Meeting
CRD Chief Technologist David Bailey will give a keynote address at the May 6 “SHARCNET Research Day,” an annual meeting of researchers affiliated with SHARCNET (Shared Hierarchical Academic Research Computing Network), Canada’s leading high-performance computing network. The event will be held at York University in Toronto. Bailey’s talk, entitled “Computing as the Third Mode of Scientific and Mathematical Discovery,” will feature some of the large-scale computations done at NERSC, and gives an overview of the many components of modern high-performance computing (hardware, software, algorithms, numerical techniques, parallelization techniques, etc.), which Bailey likens to a symphony.
NERSC to Host ScicomP, SPXXL in San Francisco on May 10–14
NERSC will host ScicomP 16, the annual meeting of the IBM HPC Systems Scientific Computing User Group, on May 10–14 at the Parc 55 Union Square Hotel in San Francisco. Scientists, engineers, and IBM staff will gather to share real-world experiences porting, tuning, and running codes on IBM supercomputers. The ScicomP meeting will be co-located with the SPXXL Summer Workshop. SPXXL is an organization with systems-oriented interests in large-scale scientific technical computing on IBM hardware. NERSC staff members Richard Gerber, Scicomp President, and David Paul, SPXXL President, are local organizers of the meetings.
Several NERSC and CRD researchers will speak at ScicomP 16:
- NERSC Director Kathy Yelick: “Programming 100,000 processors”
- CRD Director/ALD Horst Simon: “From Bits to Buildings”
- John Shalf, NERSC: “Challenges on the Road to Exascale”
- Zhengji Zhao, NERSC: “Node Mapping and the Parallel Performance of the LS3DF Code on BG/P”
- Julian Borrill, NERSC/CRD: “Cosmic Background Radiation Analysis and Modeling using the NERSC Global File System”
Free registration is available to a limited number of students or others who might not otherwise be able to afford the registration fee. Contact Richard Gerber to apply.
NERSC Systems Department Has Two Job Openings
The NERSC Systems Department has an opening for a JGI Systems Group Lead, who will lead a team of high performance computing specialists (system administrators, database administrators, and user support personnel) providing the IT infrastructure for the Joint Genome Institute (JGI) under a collaboration with the NERSC Division. See job details.
There is also an opening for an experienced System Administrator who can assist in supporting the Department of Energy’s Magellan Project. This person will help in the configuration and support of a recently deployed testbed including a 5,760-core cluster. See job details.
The Lab’s Employee Referral Incentive Program (ERIP) awards $1,000 (net) to employees whose referral of an external candidate leads to a successful hire.
Yahoo Is Accepting Proposals for Use of Cloud Computing Cluster
As part of Yahoo!’s continuing effort to advance cloud and mobile computing research, they will make a cloud computing cluster available to selected universities in the U.S. beginning in Summer 2010. The initial, two-page proposal is due on May 26, 2010. If you are interested in using the cluster, please contact Masoud Nikravesh by May 14.
Yahoo!’s M45 cluster is part of the Open Cirrus Testbed created by HP, Intel, and Yahoo!, and the cluster has been in use by selected universities since November 2007. The partition Yahoo! plans to make available will consist of roughly 400 nodes with a total of 3,200 cores, over 2 terabytes of memory, and over 1 petabyte of raw storage. The Yahoo! cluster is intended to run the Apache open source software Hadoop and Pig. Each selected university will share the partition with other selected universities. The initial duration of use is six months, potentially renewable for another six months upon written agreement.
User Interface Design@Berkeley Presentations and Demos
Joint Final Presentations for three UC Berkeley user interface courses will be held on Wednesday, May 5, 6-9 pm in the Banatao Auditorium and Kvamme Atrium in Dai Hall. Presentations start at 6 pm, posters at 7:30 pm. Students in these courses spent 15 weeks designing, prototyping, and evaluating mobile user interfaces. Student teams created applications that aid dance group rehearsals, enable basketball coaches to draw new plays, give advice to beginning rock climbers, connect emergency veterinarians, help singers find perfect pitch, and more. Food and refreshments will be provided. This event is sponsored by the EECS Department, the School of Information, the Berkeley Center for New Media, the Course Thread on Human-Centered Design, and IBM Almaden.
MSRI Presents “Making Music in the 21st Century”
The Mathematical Sciences Research Institute (MSRI) invites you to a two-part free event, “Making Music in the 21st Century,” on Friday, May 7, from 4:45 to 7:15 pm. Join us for the entire evening or attend any portion of the event. The event begins with an hour-long class (4:45-5:45 pm) in the Simons Auditorium on “Playing Music on your iPhone and iPad” with Stanford Professor Ge Wang of the CCRMA (Stanford Center for Computer Research in Music and Acoustics), creator of “Ocarina” for the iPhone and “Magic Piano” for the iPad, founder and director of SLOrk (Stanford Laptop Orchestra) and MoPhO (Stanford Mobile Phone Orchestra) and Jieun Oh, Co-Director of SLOrk and MoPhO.
A reception will be held from 5:45 to 6:15 pm in the Atrium. From 6:15 to 7:15 pm, a lecture/demonstration by Ge Wang and Jieun Oh will take place in the Simons Auditorium. They will also be joined by MSRI’s Special Projects Director, Bob Osserman, and Robert Bryant, Director of MSRI, for a discussion about new technologies in music-making.
This Week’s Computing Sciences Seminars
The Recent Status of the Japanese Next Generation Supercomputer Project
Monday, May 3, 2:30–3:30 pm, 50F-1647
Dr. Kenichi Miura, National Institute of Informatics, Tokyo, Japan
The Next Generation Supercomputer Project in Japan started in April 2006 as a national project, which is aimed at developing a 10 petaflops class supercomputing system by June 2012. The project is being carried out by RIKEN, with partners in industry, universities, and the government, under an initiative by MEXT (the Ministry of Education, Culture, Sports, Science and Technology). The elements of the project are as follows:
- Design, build, and set up the Next-Generation Supercomputer, one of the world’s fastest and most advanced computers, with a speed of 10 petaflops.
- Develop and distribute large-scale software applications (the “Grand Challenge” software) that make full use of the supercomputer. In particular, nano-science/nano-technology and life science have been identified as the high priority application areas.
- Connect the supercomputer to the HPC Infrastructure, a multilayered environment for the shared use of supercomputers across Japan using the high speed network.
- Establish a center to run the supercomputer, to be the world’s top center of excellence in the field of supercomputing.
In my talk, I will describe the recent status of the project.
Behavior of Machine Learning Algorithms in Hostile Environments
Thursday, May 6, 3:00-4:30 pm, 405 Soda Hall, UC Berkeley
Blaine Nelson, UCB EECS
Machine learning algorithms are rapidly emerging as a vital tool for data analysis and autonomic systems. Learners can infer hidden patterns in large complicated datasets, adapt to new behaviors, and provide statistical soundness to decision-making processes. This makes them useful for many emerging tasks in security, networking, and large-scale systems. Unfortunately, learning techniques also expose systems to a new class of security vulnerabilities—learners themselves can be susceptible to attacks. Many common learning algorithms were developed under the assumption that training data is from a natural or well-behaved distribution, but this assumption is perilous in a security-sensitive setting. With financial incentives encouraging ever more sophisticated adversaries, attacks increasingly target these learners, and an intelligent adversary may alter their behavior based on knowledge of the learner’s shortcomings or mislead it by cleverly crafting data to corrupt the learning process.
I provide an overview of the field and a unifying framework for categorizing threats against machine learning algorithms. Based on this framework, I describe how security principles apply to two applications: spam filtering and denial-of-service detection. In each, I develop attacks that effectively compromise existing learning techniques and defenses that make the learning more resilient against these attacks. Further, I generalize a general framework for characterizing the difficulty of query-based evasion attacks to a broader class of learning algorithms.
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.