InTheLoop | 09.13.2010
September 13, 2010
Reminder: CS All-Hands Meeting at 11:00 am Today (Monday, Sept. 13)
Computing Sciences will hold an important all-hands meeting at 11 a.m. Monday, Sept. 13, on the lawn in front of the Lab cafeteria. All CS staff are encouraged to attend this meeting, which will feature several significant announcements followed by a question-and-answer session. The meeting is expected to conclude around 11:30 am.
Starting at 11:30 am, the announcements will be made in a videoconference to Room 238 at OSF for NERSC staff in Oakland. Please plan to attend one of the two sessions. Employees who cannot attend the meeting in person or via videoconference can call into the 11:30 session by dialing 1-866-740-1260 and entering the passcode 4867377.
Complex Systems Group Formed in Computational Research Division
Mathematical research is playing an increasingly important role in the analysis and optimization of complex systems, including application areas relevant to the DOE mission such as networked interconnected systems like computer networks and electrical power grids. Berkeley Lab’s Computational Research Division (CRD) has significant research activities in this area, and several projects in CRD develop the algorithms, tools, and techniques to meet the analysis challenges posed by complex systems.
Therefore, effective September 1, a new group has been created to explore new opportunities in this field in a more focused way. The Complex Systems Group (CXG) will be an independent group within CRD’s High Performance Computing Research Department. CRD Chief Technologist David Bailey will be the acting CXG group lead. The following CRD staff and visitors will transfer from the Future Technologies and Scientific Computing groups to CXG: Aydin Buluc, Orianna Demasi, Alexander Kaiser, Sean Peisert, and David Leinweber.
ASCR Requests Nominations for SciDAC Division Director
On September 8, Daniel Hitchcock, Acting Associate Director for Advanced Scientific Computing Research (ASCR) in the DOE Office of Science, distributed a letter requesting nominations for one of two ASCR division directors. Some key excerpts:
I am writing to request your help in identifying candidates for the position of Division Director for the Computational Science Research and Partnerships (Research) Division in the Office of Advanced Scientific Computing Research (ASCR) within the U.S. Department of Energy’s (DOE’s) Office of Science. The ASCR program is the nation’s leading program devoted to research in applied mathematics, computer science, next generation high performance networks and computational science. Note that although this division has sometimes been referred to as the “SciDAC” Division, it actually includes the entire ASCR research portfolio.
I am seeking suggestions for candidates for this position, including self nominations. Each individual suggested will be sent a letter providing information on the position and how to apply and will be encouraged to submit an application. Please send your suggestions to me at Daniel.Hitchcock@science.doe.gov, with e-mail addresses for all candidates, by Tuesday, September 14th….
The Director of ASCR’s Research Division is responsible for the overall management of the programs in the division including: strategic program planning; budget formulation and execution; project management program integration with ASCR’s Facilities Division, other Office of Science activities and with the DOE technology offices; interagency and international liaison; and management of the federal and rotator technical and administrative staff in the ASCR Research Division office. For more information on the program please go to http://www.science.doe.gov/ascr/Research/Research.html….
The job announcement, which will be open until October 12, 2010, can be found here.
UCOP Is Exploring Expansion of Discovery Grant Program
The Office of Research and Graduate Studies (ORGS) at the UC Office of the President is exploring the expansion of proof-of-concept grant funding within the Discovery Grant Program.
As a first step in the planning process, ORGS is soliciting initial Letters of Interest (LOIs) from UC researchers with Principal Investigator status. Scientists at Lawrence Berkeley Laboratory are also eligible to apply. LOIs will be used to inform decision-making regarding proceeding with this program, to assess the likely demand for this funding opportunity, and to determine the scope and requirements of full proposals. Submission of this initial LOI is not a mandatory prerequisite for applying to the full Call for Applications/RFP when issued. The LOI submission window is from Friday, September 10, through Tuesday, October 26.
AIP State Department Science Fellowship Is Accepting Applications
Most of the foreign policy issues faced by the U.S. Department of State have a scientific or technical component. The American Institute of Physics State Department Science Fellowship is intended to enhance the S&T capacity of the Department by enabling at least one scientist annually to work at the Department’s Washington, DC headquarters for a one-year term.
This is a unique opportunity for a scientist to contribute scientific and technical expertise to the Department and raise awareness of the value of scientific input. In turn, scientists broaden their experience by interacting with policymakers in the federal government and learning about the foreign policy process. The application deadline is November 1. Go here for more information.
BLI Offers Class on Enhancing Research Presentations
What’s the most effective way to present your research? On Sept. 21, the Berkeley Lab Learning Institute will offer a class on “Creating Effective Research Presentations” (BLI0116). This two-hour workshop, designed for postdocs, research associates, and scientists, provides tips on designing effective slides and avoiding common presentation mistakes. The course instructor is director of UC Berkeley’s Technical Communications Program. Registration is required.
Invitation to a Book Launch Party for The Two Body Problem
There will be a book launch party for The Two Body Problem by author/UC professor Thomas Farber and mathematician/filmmaker/UC professor Edward Frenkel at 7:30 pm on Wednesday, September 15, at La Pena Cultural Center, 3105 Shattuck Avenue in Berkeley. This event, which is open to the public, will include the screening of a conversation/interview of the authors by the publisher Andrea Young, Q & A, and book signing by the authors. Go here for details.
In The Two-Body Problem — written as a screenplay rather than a novel — two Americans, a writer and a mathematician, happen to meet in a beach town in the South of France. Girl-watching, the men trade stories about bachelor life, relationships, loves lost. In the cafe and on the beach, they talk about how their vocations shape their views of truth and beauty as well as the pursuit of sex and love. Encountering several young women as the day passes, they share with them anecdotes about art and mathematics discussed playfully, passionately, seductively. That evening, the two men and the women met earlier are partying when a surprising guest shows up.
This Week’s Computing Sciences Seminars
Large Deviations and Application to Fine TCP Modeling
Monday, Sept. 13, 1:30–2:30 pm, 400 Cory Hall (Hughes Room), UC Berkeley
Patrick Loiseau, INRIA Paris-Rocquencourt
A large proportion of today's Internet traffic is carried using the TCP transport protocol. Yet, most of the numerous models that have been proposed in the last decade to understand the performance achieved by one or several long TCP connections are limited to the prediction of a mean throughput value. In this talk, we expose a new model, based on an original large-deviations theorem, which allows finely predicting the deviations of the throughput around its mean.
We first present a new large-deviations theorem valid on almost-every realization of a stationary mixing process. This theorem expresses the "ergodic transfer" of classical large-deviations properties of a process to almost-every realizations. It allows estimating the proportion of time, within a single realization averaged at a given (large) scale, where the process deviates from its almost-sure mean value.
Then, applying this theorem to a Markov chain modeling the TCP congestion-window evolution, we show that it permits in practice to accurately quantify and to statistically bound the variations of the throughput within a long flow. The Markov-chain model can take into account various network conditions. We discuss in particular the classical case of Bernoulli losses, as well as more realistic cases from experiments and real Internet traces. We finally discuss extensions of the model to multiple competing TCP connections and to TCP variants other than Reno.
Anton: A Special-Purpose Machine That Achieves a Hundred-Fold Speedup in Biomolecular Simulations
Tuesday, Sept. 14, 11:00 am–12:00 pm, Wozniak Lounge, Soda Hall, UC Berkeley
David E. Shaw, D. E. Shaw Research and Center for Computational Biology and Bioinformatics, Columbia University
Molecular dynamics (MD) simulation has long been recognized as a potentially transformative tool for understanding the behavior of proteins and other biological macromolecules, and for developing a new generation of precisely targeted drugs. Many biologically important phenomena, however, occur over timescales that have previously fallen far outside the reach of MD technology. We have constructed a specialized, massively parallel machine, called Anton, that is capable of performing atomic-level simulations of proteins at a speed roughly two orders of magnitude beyond that of the previous state of the art. The machine has now simulated the behavior of a number of proteins for periods as long as a millisecond — approximately 100 times the length of the longest such simulation previously published — revealing aspects of protein dynamics that were previously inaccessible to both computational and experimental study. The speed at which Anton performs these simulations is in large part the result of a tightly coupled codesign process in which the machine architecture was developed in concert with novel algorithms, including an asymptotically optimal parallel algorithm (with highly attractive constant factors) for the range-limited N-body problem.
Using Memory Traffic Predictions to Estimate the Runtime of Linear Algebra Kernels
Wednesday, Sept. 15, 1:00–2:00 pm, 50F-1647
Ian Karlin, University of Colorado at Boulder
Data movement limits the performance of many scientific computing applications. For these programs, runtimes are most accurately expressed in terms of memory traffic, not floating-point operations. To significantly improve the performance of these applications, memory traffic must be reduced. Tuning techniques such as loop fusion decrease data movement, often producing speedups proportional to the resulting reduction in memory accesses. However, loop fusion sometimes decreases performance by causing capacity misses in caches and registers. Whether fusion causes misses depends on hardware and routine characteristics. Finding the optimal amount of fusion, which requires trying all possible fusion strategies for all sizes of interest, is often infeasible.
In this talk, we present a memory model that predicts the performance of linear algebra kernels. We include only the most distinguishing machine and routine features, allowing for an economical comparison while maintaining accuracy. The model works in two phases, first calculating the amount of data accessed from each memory structure for a calculation. We then turn these memory traffic predictions into runtime estimates and establish their accuracy for both single processor and shared memory parallel systems. Our model is integrated into a compilation framework where its runtime estimates reduce the large number of versions of a routine to a practical collection that is practical to test. Through the use of the model, compile time is greatly reduced without sacrificing routine speed.
Lightweight, Deterministic Concurrency and Preemption in C and Java
Wednesday, Sept. 15, 4:00–5:00 pm, 540 Cory Hall, UC Berkeley
Reinhard von Hanxleden, University of Kiel
The control flow of reactive systems typically entails not just the sequential control flow found in traditional programming languages, such as conditionals and loops, but also exhibits concurrency and preemption. How to express this adequately in an imperative language such as C or Java is a notoriously difficult problem. Threads and associated synchronization primitives are widely used, but entail significant overhead and easily lead to non-determinism and deadlock.
I will present a lightweight approach to embed reactive, deterministic control flow in traditional sequential imperative programming languages. The approach is inspired by coroutines and the synchronous model of computation. A prototypical implementation for C, called Synchronous C (SC), is freely available. An adaptation for Java is currently under development.
Link of the Week: God’s Number Is 20
With about 35 CPU-years of idle computer time donated by Google, a team of researchers has essentially solved every position of the Rubik’s Cube and shown that no position requires more than 20 moves. The most efficient algorithm, one that always uses the shortest sequence of moves, is known as God’s Algorithm. The number of moves this algorithm would take in the worst case is called God’s Number.
It took fifteen years after the introduction of the Cube to find the first position that provably requires 20 moves to solve; it is appropriate that fifteen years after that, the researchers proved that 20 moves suffice for all positions.
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.