A-Z Index | Phone Book | Careers

InTheLoop | 03.09.2015

March 9, 2015

ESnet’s Science DMZ is Foundation for New Infrastructure Linking California’s Top Research Institutions

The Pacific Research Platform, a cutting-edge research network infrastructure based on ESnet’s Science DMZ architecture, will link together the Science DMZs of dozens of top research institutions in California. The Pacific Research Platform was announced Monday, March 9, at the CENIC 2015 Annual Conference.

The new platform will link the sites via three advanced networks: the Department of Energy’s Energy Science Network (ESnet), CENIC’s California Research & Education Network (CalREN) and Pacific Wave. Initial results for the new infrastructure will be announced in a panel discussion during the conference featuring Eli Dart (ESnet), John Haskins (UC Santa Cruz), John Hess (CENIC), Erik McCroskey (UC Berkeley), Paul Murray (Stanford), Larry Smarr (Calit2), and Michael van Norman (UCLA). The presentation will be »live-streamed at 4:20 p.m. Pacific (Daylight Savings) Time on Monday, March 9. »Read more.

CCSE Research in Hydrodynamics Featured on Office of Science Highlights Page

An article on research by members of CRD’s Center for Computational Sciences and Engineering to model fluids at the mesoscale is a science highlight on the DOE Office of Science website. Mesoscale refers to the regime between the microscale, where physicists and chemists use molecular dynamics to characterize the motion of atoms and molecules, and the macroscale, where scientists and engineers apply continuum models to describe fluid flow that is observable with the naked eye. Read the full story. »Read the full story.

NERSC Users Learn Code Optimization Tips and Tricks at First Hackathon

NERSC users got the opportunity to optimize a computationally intensive piece of code at a one-day hackathon held February 25 during the annual meeting of NUG (NERSC Users Group). The exercise was conceived as a way to help researchers adapt their scientific codes to run well on Cori, the Intel Xeon Phi processor-based system slated to arrive at NERSC in 2016. During the hackathon, NERSC staff were on hand to provide optimization strategies, assist with coding tools and answer questions. Attendees did the work of refactoring code to make it run faster. Remote attendees were able to follow along and pose questions via a chat interface. »Read more.

ESnet Taking perfSONAR to the Heartland

As part of its member training program, the Great Plains Network is hosting a one-hour webinar on "perfSonar and Network Monitoring" led by Jason Zurawski of ESnet’s Science Engagement team. The March 13 workshop is open to members of the Great Plains Network, a consortium of universities in Arkansas, Iowa, Kansas, Minnesota, Missouri, Nebraska, Oklahoma and South Dakota.

Berkeley Lab Well-Represented at Tapia Diversity in Computing Conference

At the ACM Richard Tapia Celebration of Diversity in Computing Conference, held February 18-21 in Boston, staff from Lawrence Berkeley National Laboratory’s (Berkeley Lab) Computing Sciences organization delivered talks, participated in mentoring and did recruiting. And for the ninth consecutive conference, Berkeley Lab was a contributing supporter.

Elizabeth Bautista of the National Energy Research Scientific Computing Center, in collaboration with staff from Lawrence Livermore and Oak Ridge national laboratories, led a mentoring breakfast sponsored by the Department of Energy’s Advanced Scientific Computing Research office. About 130 graduate students attended the session. Bautista also presented at two professional development workshops. Silvia Crivelli of the Computational Research Division and Sowmya Balasubramanian of ESnet participated on the panel “Multidisciplinary Computer Science: Solving the Important Questions of Today and Tomorrow” and the Birds of a Feather (BoF) session “Exploring Cybersecurity Experimentation with Linux and DETER.” Computing Sciences recruiter Jeff Todd staffed a table, talking with attendees about student and career opportunities at Berkeley Lab. »Read more.

Wehner on Extreme Weather Events in Ars Technica

A recent article in the online magazine Ars Technica features the work of computational climatologist Michael Wehner, senior staff scientist with CRD. The article explores his groups' work modeling extreme weather events and calculating the probable impact of climate change. "When an extreme weather event happens, the public wants to know—is this climate change?" Wehner is quoted asking at the 2015 annual AAAS meeting held in San Jose, Calif. last month. The article continues

[S]cience is on solid ground when it examines weather events in terms of probabilities: is the risk of a given event higher? Will the magnitude of a given type of event change? Wehner went through some historic events and examined how climate change shifted these probabilities. For example, events similar to Europe's 2003 heat wave (which saw 70,000 deaths) are already twice as likely to occur given the amount we've warmed over preindustrial conditions.

The article goes on to discuss the work of Wehner and other scientists studying climate change and future weather conditions. »Read more.

This Week's CS Seminars

»CS Seminars Calendar

ExaFMM -- a Testbed for Comparing Various Implementations of the FMM

Monday, March 9, 2–3 p.m., Bldg. 50F, Room 1647
Rio Yokota, Extreme Computing Research Center, King Abdullah University of Science and Technology (KAUST)

The software design space for the fast multipole method (FMM) is large, where various schemes exist for partitioning, communication, tree traversal, and translation operators. The optimal choice is problem and architecture dependent, and most FMM codes focus on a specific combination of the two. ExaFMM is a open source FMM code that provides the capability to explore this vast design space without sacrificing speed, scalability, or readability.

Applied Math: Phase-field models for multiphase complex fluids: modeling, numerical analysis and simulations

Wednesday, March 11, 2:30–3:30 p.m., 939 Evans Hall, UC Berkeley Campus
Jie Shen, Purdue University

I shall present some recent work on phase-field model for multiphase incompressible flows. We shall pay particular attention to situations with large density ratios as they lead to formidable challenges in both analysis and simulation.

I shall present unconditionally energy stable, decoupled numerical schemes which only require solving a sequence of linear elliptic equations at each time step for solving this coupled nonlinear system, and show ample numerical results which not only demonstrate the effectiveness of the numerical schemes, but also validate the flexibility and robustness of the phase-field model.

Scientific Computing supporting Photon and Neutron Facilities at STFC

Thursday, March 12, 10–11 a.m., Bldg. 70A, Room 3377
Brian Matthews, Scientific Computing Department, Science and Technology Facilities Council, United Kingdom

The Scientific Computing Department of STFC provides advanced computing technology to the UK’s science community.  A major activity of the department is to provide infrastructure and software to the neutron and photon sources supported by STFC, including the ISIS neutron spallation source and Diamond synchrotron. 

In this talk, I shall introduce the range of work of the department from high performance computing and data infrastructure services through to developing simulation applications.  I shall then discuss our approach to data management and access of the data from large scale facilities to support experiments, in particular the ICAT catalogue.  I shall describe some initiatives to extend and integrate the infrastructure in response to the increasing capability of new instruments.  I shall go on to discuss how we have been collaborating with partners around Europe and elsewhere to provide common data infrastructure across facilities, within the PanData consortium and now in the emerging PanDaas initiative.

Efficient numerical methods for micro/nanoscale heat conduction

Thursday, March 12, 11 a.m.–12 p.m., Bldg. 50F, Room 1647
Jean-Philippe Peraud, Massachusetts Institute of Technology

We present efficient numerical methods for solving the Boltzmann transport equation for phonons with a particular emphasis on Monte Carlo methods. We show that variance reduction can be achieved by simulating particles that represent the deviation from a given equilibrium, resulting in speedups of several orders of magnitude compared to direct simulation Monte Carlo methods applied to phonon transport. This control variates approach removes the computational cost associated with the statistical uncertainty of equilibrium fluctuations, and therefore captures arbitrarily small deviations from equilibrium. In addition, we show that this method alleviates the stiffness associated with the continuum limit by adaptively focusing the computational effort in the regions where kinetic effects prevail.

We further develop the deviational approach by showing that when the Boltzmann equation can be linearized with respect to the chosen equilibrium distribution, trajectories of the deviational particles can be simulated independently. This yields an efficient and simple algorithm which does not rely on time discretization and which proves particularly useful for applications of interest such as 3D simulations of transient heat conduction problems or thermal conductivity calculations. It is also very easily parallelizable and enables the use of adjoint-based methods for achieving high levels of resolution.

We will conclude with a discussion of asymptotic methods for solving the low Knudsen limit. This analysis quantifies the deviations from Fourier solutions arising due to the inhomogeneity introduced by the system boundaries. The asymptotic solution shows that the Fourier description can be still used in the presence of small to moderate size effects because deviations from Fourier can be rigorously captured by applying modified boundary conditions to the heat equation and superposing boundary layer corrections to the resulting solution in the vicinity of the boundaries. We show that the modified boundary conditions are of the temperature jump type and that the jump functional form and associated boundary layer corrections are uniquely determined from the material properties and material-boundary interaction law.

A Parallel Orbital-updating Based Optimization Approach for Electronic Structure

Friday, March 13, 10 a.m.–11 a.m., Bldg. 50B, Room 2222
Aihui Zhou, Chinese Academy of Sciences

In this presentation, we will talk about a parallel orbital based optimization method for electronic structure calculations. This approach is a generalization of the parallel orbital-updating approach for solving the single-particle nonlinear eigenvalue problems to the associated energy minimization problems. With this approach, the solution of the direct minimization problem is reduced to some solutions of independent local minimization problems based single orbital searches and a small scale global minimization problem. It is demonstrated by our numerical experiments that this  approach is quite efficient for electronic structure calculations for a class of molecular systems. This presentation is based on some joint work with Xiaoying Dai, Xingao Gong, Zhuang Liu, and Xin Zhang.