InTheLoop | 01.24.2011
January 24, 2011
Engineering Mixed Traffic on ESnet’s ANI Testbed
The first crop of experiments using ESnet’s Advanced Networking Initiative testbed are now in full swing. In a project funded by the DOE Office of Science, Prof. Malathi Veeraraghavan and post-doc Zhenzhen Yan at the University of Virginia, along with consultant Prof. Admela Jukan, are investigating the role of hybrid networking in ESnet’s next generation 100 Gbps network. Their goal is to learn how to optimize a hybrid network comprised of two components, an IP datagram network and a high-speed optical dynamic circuit network, to best serve users’ data communication needs. Read more.
Contra Costa College Parallel Computing Club Visits NERSC
Eleven members of the Contra Costa College Parallel Computing Club paid a visit to NERSC at Berkeley Lab’s Oakland Scientific Facility on January 14. The students, along with instructor Tom Murphy, were welcomed by NERSC’s User Services Group Lead, Katie Antypas, who gave an introduction to parallel computing and NERSC as well as a tour of the facility’s computer room. The visit grew out of a conversation at the annual Supercomputing Conference held last November in New Orleans.
MSRI Mathematics Festival for Youngsters Sunday at UC Berkeley
Middle school and high school students can take part in fun and engaging mathematical activities at the upcoming Julia Robinson Mathematics Festival on Sunday, January 30 at UC Berkeley, sponsored by the Mathematical Sciences Research Institute (MSRI). There will be a morning session from 8:45 am to 1:15 pm and an afternoon session from 12:15 to 4:30 pm. (The overlap is a public lecture about the mathematics of the Rubik’s cube.) The festival will take place in the Pauley Ballroom in the MLK Jr. Student Union on the UC Berkeley campus at 2475 Bancroft Way (where Telegraph Avenue meets the campus). The event is free and open to the public, but please register in advance to assist in planning.
This Week’s Computing Sciences Seminars
Energy Efficient Computing
Monday, January 24, 1:00–2:30 pm, 50B-4205
Tajana Šimunić Rosing, University of California, San Diego
In this talk I give an overview of the approaches we have developed at UCSD to significantly lower the energy consumption in computing systems. We derived optimal power management strategies for stationary workloads. Run-time adaptation can be done via an online learning algorithm that selects among a set of policies. We generalize the algorithm to include thermal management, since we found that minimizing the power consumption does not necessarily reduce the overall energy costs. To reduce the performance costs typically associated with state of the art thermal management techniques, we developed a new set of proactive management policies. The experimental results using datacenter workloads show that our proactive technique is able to dramatically reduce the adverse effects of temperature by over 60%. Further extending our work in datacenter energy management, we have shown that symbiotic scheduling of workloads in virtualized environments can lead to average 15% energy savings with 20% performance benefit in high utilization scenarios. Integrating SLA management capabilities along symbiotic workload scheduling enables 2x or higher increases in energy efficiency while meeting the SLA requirements. This is done by running a mix of SLA sensitive workloads and batch jobs per server. Going forward we are working on estimating the energy cost of communication at local and global scales, and plan to use this information for opportunistic virtual machine balancing both within and across geographically distributed datacenters.
Experiences Evolving a New Analytical Platform: What Works and What’s Missing
Tuesday, January 25, 3:00–4:00 pm, 50F-1647
Jeff Hammerbacher, Cloudera
At Cloudera, we augment existing analytical platforms with some new tools for data management and analysis. In this talk, we’ll share some experiences of what has worked across industries and workloads, and what new software components might help complete a new analytical platform.
Fragment-HMM: A New Approach to Protein Structure Prediction
Wednesday, January 26, 10:00–11:00 am, 54-130B (Pers. Hall Addition)
Shuai Cheng Li, University of California, Berkeley (Alvarez Fellowship Candidate)
Protein structure prediction has been a heuristic science. From Monte Carlo fragment assembly, decoy clustering, selection, to refinement and consensus, there is no unified model or theory governing the complete process. We present a new idea that results in a unified and powerful theory for this problem. In particular, we design a simple position-specific hidden Markov model to model the protein folding process. Our new framework naturally repeats itself to converge to a final target, conglomerating fragment assembly, clustering, target selection, refinement, and consensus, all in one process. Our initial implementation of this theory converges to within 6A of the native structures for 100% of decoys on all six standard benchmark proteins used in ROSETTA, which achieved only 14% to 94% for the same data. The qualities of the best decoys and the final decoys our theory converges to are also notably better. In addition, the project is extended to combine chemical shift data to determine protein structures. (Based on joint work with Dongbo Bu, Jinbo Xu, and Ming Li.)
Wednesday, January 26, 4:00–5:00 pm, 939 Evans Hall, UC Berkeley
Vladimir Rokhlin, Yale University
No abstract available.
Wednesday, January 26, 6:00–8:00 pm, OSF 943-238
John Levesque, Cray, Inc.
No abstract available.
GPU-Accelerated Molecular Dynamics Models for Studying Soft Matter Systems
Thursday, January 27, 10:00–11:00 am, 50B-4205
Carolyn Phillips, University of Michigan (Alvarez Fellowship Candidate)
Many properties of materials are determined by micro- to nanoscale features. And materials at the scale of tens to hundreds of atoms evidence radically different physical properties than their bulk counterparts. Already, a new class of nano-engineered materials is being developed whose properties are carefully controlled at the molecular level. In the future, nano-engineered materials will not be constructed top-down, component by component, but, rather, self-assembled from the bottom-up through a careful choice of specifically designed nanoscale components. Even coarse-grained models of nanoparticle self-assembly can be computationally expensive. To observe the critical features, many systems require thousands of particles and tens of millions of time steps, and have large parameter spaces to explore. Developed in my research group, the GPU accelerated HOOMD-Blue, Highly Optimized Object-oriented Many-particle Dynamics–Blue Edition, performs general-purpose particle dynamics simulations on a single GPU-enabled workstation, but achieves the performance of dozens of processor cores. In my research in soft matter self-assembly, I use GPU-accelerated molecular dynamics to rapidly generate and explore phase diagrams of polymer-tethered nanoparticles. I am investigating new computationally efficient models for capturing the self-assembly behavior of heterogeneous and anisotropic nanoscale particles found in soft matter systems. These particles will be the building blocks of novel materials designed to self-assemble spontaneously. (Note: This is an updated version from September 2010 given by Ms. Phillips at LBNL.)
Simulation Research for Low Energy Buildings
Thursday, January 27, 12:00–1:00 pm, 90-4133
Wangda Zuo, Environmental Energy Technologies Division, LBNL
Buildings consume 40% of primary energy in the U.S. This seminar introduces our research in simulation tools for low energy buildings. The goal is to reduce energy consumption and peak power demand while maintaining or improving the indoor environment quality.
To quickly provide spatial and temporal airflow information for ventilation system design and indoor environment control, we have applied and further developed a fast fluid dynamics (FFD) model, which is 50 times faster than computational fluid dynamics (CFD). By running the FFD simulation in parallel on a graphics processing unit, it can be 1500 times faster than CFD on a CPU.
To enable rapid prototyping of new HVAC systems and to support the design and performance assessment of advanced control systems, we have developed an open-source Modelica library for building energy and control systems. By separating physical modeling from numerical solution, our approach can make modeling flexible and fast to accelerate innovation of low energy systems.
In addition, the coupling of FFD on GPU and Modelica will be useful for the design and control of HVAC system for indoor environment with non-uniform air distribution, such as auditoriums, data centers and buildings using natural ventilation.
Silicon Photonic Devices for On-Chip Microsystems
Friday, January 28, 1:00–2:00 pm, 521 Cory Hall (Hogan Room), UC Berkeley
Zhiping (James) Zhou, Peking University and Georgia Institute of Technology
Silicon photonics has a potential as an efficient and low cost optical solution for high density data communications in optical fiber systems and computer systems. It is expected that successful monolithic integration of silicon based photonic devices and microelectronic devices will lead to a more significant “micro-optoelectronics revolution” than the well-known “microelectronics revolution”. From this point of view, this talk will present recent development in silicon based photonic devices and efforts in large scale integration. Along this line, our own research on silicon photonic devices and their application for on-chip microsystems will also be presented.
Link of the Week: Algorithms Take Control of Wall Street
Algorithms have become so ingrained in our financial system that the markets could not operate without them. The result is a system that is more efficient, faster, and smarter than any human. It is also harder to understand, predict, and regulate. At its best, this system represents an efficient and intelligent capital allocation machine, a market ruled by precision and mathematics rather than emotion and fallible judgment. But at its worst, it is an inscrutable and uncontrollable feedback loop. For better or worse, the computers are now in control. Read more in Wired.
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.