InTheLoop | 10.20.2014
ESnet to Boost Big Data Transfers by Extending 100G Connectivity across Atlantic
DOE’s Energy Sciences Network, or ESnet, is deploying four new high-speed transatlantic links, giving researchers at America’s national laboratories and universities ultra-fast access to scientific data from the Large Hadron Collider (LHC) and other research sites in Europe. ESnet’s transatlantic extension will deliver a total capacity of 340 gigabits-per-second (Gbps) and serve dozens of scientific collaborations. To maximize the resiliency of the new infrastructure, ESnet equipment in Europe will be interconnected by dedicated 100 Gbps links from the pan-European networking organization GÉANT.
Funded by the DOE’s Office of Science and managed by Berkeley Lab, ESnet provides advanced networking capabilities and tools to support U.S. national laboratories, experimental facilities and supercomputing centers. »Read more.
CRD Scientists Address U.S., European Big Data Cooperation at French Embassy
On October 9, members of Berkeley Lab's Computational Research Division participated in an invitation-only "Big Data for Society Workshop" held at the French Embassy in Washington, D.C. The goal of the workshop was to explore science and technology perspectives and potential collaborations between Europe—France in particular—and the U.S. on big data.
Berkeley Lab scientists David Brown and Peter Nugent delivered two invited keynote speeches. Division Director Brown's keynote was entitled "Societal Impacts of Data-Intensive Science Research and Development in the U.S." Nugent, a senior staff scientist and CRD deputy for scientific engagement, addressed "Data Analysis in Astronomy."
One round-table explored the point of view of U.S. funding agencies concerning the participation of U.S. researchers in European Projects, in particular Horizon 2020 (H2020); another explored the point of view of European Union members about EU-US partnerships in H2020.
The meeting was organized by Marc Daumas, the Attaché for Science and Technology at the French Embassy. Cyrus Wadia, Co-Director of UC Berkeley Haas Energy Institute and guest scientist at LBNL, currently on detail at the OSTP (Office of Science and Technology Policy), White House, helped Marc prepare a list of invitees. The meeting resulted from discussions involving personnel from IRIT and LBNL, and was attended by more than 60 people, including participants from U.S. funding agencies, Science Attachés from embassies of the E.U. member states, and researchers from France and the U.S. A follow-up meeting is in the works, with the purpose of identifying concrete research partnerships. »Learn more.
Katie Antypas Wows Audience at '8 Big Ideas' Talks
Oakland North, an online news project of the Berkeley Graduate School of Journalism, and the San Francisco Chronicle focused on NERSC Services Department Head Katie Antypas in their coverage of the Lab's recent "Science at the Theater" event. At the event Antypas explained the energy barrier facing the advancement of supercomputing and how NERSC's Cori system addressed that barrier, "with the charisma of a TED Talks performer." The article is extracted below and available on »Oakland North and »San Francisco Chronicle websites. You can also »watch Antypas' performance on Youtube.
- Katie Antypas strode across the stage at Kaiser Center in downtown Oakland before an audience composed of scholars, kids and residents. “What happens when you are sitting at home, you open up your laptop?” she asked. “Your laptop starts to get hot.”
- As the first speaker at the “8 Big Ideas” event, Antypas announced a supercomputer system called Cori, which will come online in 2016. During the talk, she broke the complicated idea into understandable daily situations and digestible concepts with the charisma of a TED Talks performer.
- Antypas is the service department head for the National Energy Research Scientific Computing Center and one of the eight speakers at the “8 Big Ideas” event last Wednesday, which was hosted by the Lawrence Berkeley National Laboratory as part of its “Science at the Theater” initiative. During the event, eight scientists were invited to present game-changing concepts and progressive ideas in eight minutes each. »Read more.
This Week's CS Seminars
Stochastic Reaction-Diffusion Methods for Modeling Cellular Processes
Wednesday, Oct. 22, 3:30 p.m. – 4:30 p.m., 939 Evans Hall, UC Berkeley
Samuel Isaacson, Boston University
Particle-based stochastic reaction diffusion methods have become a popular approach for studying the behavior of cellular processes in which both spatial transport and noise in the chemical reaction process can be important. While the corresponding deterministic, mean-field models given by reaction-diffusion PDEs are well-established, there are a plethora of different stochastic models that have been used to study biological systems, along with a wide variety of proposed numerical solution methods.
In this talk I will introduce our attempt to rectify the major drawback to one of the most popular particle-based stochastic reaction-diffusion models, the lattice reaction-diffusion master equation (RDME). We propose a modified version of the RDME that converges in the continuum limit that the lattice spacing approaches zero to an appropriate spatially-continuous model. I will then discuss some application areas to which we are applying these methods, focusing on how the complicated ultrastructure within cells, as reconstructed from X-ray CT images, might influence the dynamics of cellular processes.
FASTMath: What the numerical library developers require from future programming environments
Friday, Oct. 24, 11 a.m. – 12:30 p.m., 50A-5132
Sherry Li, Chao Yang, Ann Almgren, Brian Van Straalen, Anshu Dubey and Katie Antypas, LBNL
The FASTMath SciDAC Institute develops and deploys scalable mathematical algorithms and software tools for reliable simulation of complex physical phenomena and collaborates with application scientists to ensure the usefulness and applicability of FASTMath technologies. The project consists of some of the top advanced numerical library developers in the DOE. On October 1-2, the SciDAC team got together at Argonne National Laboratory to discuss the FASTMath strategy for next generation architectures, particularly those with heterogeneous nodes. The meeting included presentations from the LCFs and NERSC on their long-term plans for their hardware acquisitions followed by short presentations from the software teams in FASTMath to hear the lessons learned, war stories, etc from their experiences to date in developing and using these machines and portable programming abstractions. The following day included breakout sessions with detailed discussions about possible strategies for moving forward, who to handle interoperability across the software stack, what we need from the facilities, etc.
In this exascale seminar, the participants in the FASTMath meeting will recap their input to the meeting and highlights and provide insight into programming environment requirements for future generations of advanced numerical libraries.