InTheLoop | 09.30.2013
Crane to Take CRT Construction to the Next Level
The Lab skyline got a dramatic new addition over the weekend when a tall yellow crane was moved into place alongside the under-construction CRT Facility. With all the concrete poured for the ground floor of the building, the crane will be used to move the steel and glass components onto the structure to build the two floors for staff. You can catch glimpses of the crane via the CRT webcam view.
SC13: Stage a Demo, Host a Roundtable Discussion in DOE Booth
At the SC13 conference in Denver, DOE is sponsoring a joint booth featuring the expertise, facilities and accomplishments of 15 national labs. The booth is created on the pattern and structure developed by LBNL over the past five years, featuring an open design with chairs and tables to encourage interactions.
A limited number of opportunities are available for Berkeley Lab researchers who want to demonstrate new technologies or programs or host a one-hour roundtable discussion on a topic of broad interest. If you or your group is interested in taking advantage of either opportunity, please contact Jon Bashor (email@example.com) by Monday, Oct. 7, for details.
Q&A's: Kathy Yelick, Sudip Dosanjh, Erich Strohmaier
In conjunction with DOE Public Affairs’ communications focus on supercomputing during the month of September, three HPC leaders from Computing Sciences shared their perspectives in Q&As.
Associate Lab Director Kathy Yelick talks about her job, how to recruit more students into the STEM fields, her own career path and her near-term view of HPC. »Read the Q&A.
NERSC Director Sudip Dosanjh talks about his impressions of NERSC, defines a supercomputer and describes the rewards and challenges of his job. »Read the Q&A.
Erich Strohmaier, head of CRD’s Future Technologies Group, is better-known for his role in launching and maintaining the twice-yearly TOP500 list of the world’s top supercomputers. DOE’s Ben Dotson posed 10 questions to Strohmaier about the origin of the list, its value to the community, the road (and speedbumps) to exascale and other topics. »Read the Q&A.
ESnet’s Inder Monga Helps Organize, Run Workshop on Terabit Networking
Although ESnet is the nation’s fastest network for science with its 100 gigabit-per-second backbone, the science networking community is already looking ahead to terabit networks. Last week, the National Science Foundation sponsored a workshop on “Scaling Terabit Networks: Breaking Through Capacity Barriers and Lowering Cost with New Architectures and Technologies.” ESnet Chief Technologist Inder Monga was a member of the workshop organizing team and also led three breakout sessions in the areas of Key Trends, Obstacles & Opportunities; Metrics, Targets, & Capabilities; and Research Priorities & Requirements.
“The meeting was well attended by a cross-section of people involved with optical components, systems and data networking folks with representatives from academia,vendors and the DOE community,” Monga said. “There were some great ideas discussed that will be written up in a report and shared within a month.”
According to materials developed to support the workshop, networks and data centers that support scientific research are experiencing exponential traffic growth, and so research backbones such as ESnet and regional networks are expected to be early adopters for terabit-scale networks. On the commercial side, video production, digital healthcare, warehouse-scale computing and latency-driven financial applications will also benefit from this new capability.
This Week’s Computing Sciences Seminar
The GNAT method for model reduction of nonlinear dynamical systems
Wednesday, Oct. 2, 3:30 - 4:30 p.m., 939 Evans Hall - UC Berkeley Campus
Kevin T. Carlberg, Sandia National Laboratories, Livermore
Abstract: Time-critical applications for systems governed by dynamical systems—such as control, fast-turnaround design, and uncertainty quantification—often demand the accuracy provided by large-scale computational models, but cannot afford their computational cost. To mitigate this bottleneck, researchers have developed model-reduction techniques that decrease the dimension of the dynamical system while preserving its key features. Such methods are effective when applied to specialized problems such as linear time-invariant systems (e.g., balanced truncation). However, model reduction for nonlinear dynamical systems has been primarily limited to methods based on the proper orthogonal decomposition (POD)–-Galerkin approach, which lacks `discrete optimality' and leads to unstable responses in many cases.
In this talk, I will present the Gauss–Newton with approximated tensors (GNAT) nonlinear model-reduction method. This method is discrete optimal, is equipped with an error bound, and leads to highly accurate responses for practical problems across a wide range of physics. I will also describe the `sample mesh' concept, which enables a practical, distributed, computationally efficient implementation of GNAT in computational-mechanics codes. Finally, I will present results for the method applied to a validated CFD model (with over 17 million unknowns) of a compressible, turbulent flow problem. Results illustrate GNAT’s favorable performance compared with other model-reduction techniques; it achieves speedups exceeding 350 with errors below 1 percent. (Joint work with Charbel Farhat, Julien Cortial and David Amsallem.)