InTheLoop | 08.19.2013
Computing Sciences Picnic this Friday at Tilden Park
All Computing Sciences staff members are invited to attend a picnic at Tilden Park from 11:30 a.m. to 2:30 p.m. on Friday, Aug. 23. The event will be held at the Padre Picnic Area on South Park Drive, just over the hill from Berkeley Lab. Food and drinks will be provided. RSVP to Leah Temple.
Jeff Broughton Named NERSC Division Deputy for Operations
Jeff Broughton has been named as the new NERSC Division Deputy for Operations. The announcement was made Aug. 15 by NERSC Division Director Sudip Dosanjh. “Rather than this being a new position, the Division Deputy title is a fitting recognition of the duties and responsibilities Jeff has taken on since he joined NERSC four years ago,” Dosanjh wrote in his announcement. “When he was hired as head of the Systems Department in 2009, Jeff brought 30 years of HPC and management experience to the position.” »Read more.
This Week’s Computing Sciences Seminars
The Berkeley Data Analytics Stack: Present and Future
Monday, August 19, 1 – 2 p.m, Bldg. 50B, Room 4205
Michael J. Franklin, University of California, Berkeley
Abstract: The Berkeley AMPLab was founded on the idea that the challenges of emerging Big Data applications require a new approach to analytics systems. Launching in early 2011, the project set out to rethink the traditional analytics stack, breaking down technical and intellectual barriers that had arisen during decades of evolutionary development. The vision of the lab is to seamlessly integrate the three main resources available for making sense of data at scale: Algorithms (such as machine learning and statistical techniques), Machines (in the form of scalable clusters and elastic cloud computing), and People (both individually as analysts and en masse, as with crowdsourced human computation). To pursue this goal, we assembled a research team with diverse interests across computer science, forged relationships with domain experts on campus and elsewhere, and obtained the support of leading industry partners and major government sponsors. The lab is realizing its ideas through the development of a freely-available Open Source software stack called BDAS: the Berkeley Data Analytics Stack. In the nearly three years the lab has been in operation, we've released major components of BDAS. Several of these components have gained significant traction in industry and elsewhere: the Mesos cluster resource manager, the Spark in-memory computation framework, and the Shark query processing system. BDAS shows up prominently in many industry discussions of the future of the Big Data analytics ecosystem - a rare degree of impact for an ongoing academic project. Given this initial success, the lab is continuing on its research path, moving "up the stack" to better integrate and support deep machine learning and to make people a full-fledged resource for making sense of Big Data.
In this talk, I'll describe the current state of BDAS with an emphasis on the key components listed above and will address our current efforts on machine learning scalability and ease of use, and hybrid human/computer processing.
Multi-Scale Modeling of Tungsten under Helium Irradiation
Thursday, August 22, 10 – 11 a.m. Bldg. 50F, Room 1647
Thibault Faney, Dept. of Nuclear Engineering, University of California, Berkeley
Abstract: In fusion reactors, plasma facing components (PFC), and in particular, the divertor will be made of tungsten and irradiated with high fluxes (up to 1023 m-2s-1) of low energy (~ 100 eV) helium and hydrogen ions. The aim of this talk is to present a framework to understand and predict defect production and diffusion, clustering and interaction close to the inner surface of the divertor due to high flux, low energy helium irradiation. The framework presented is based on a multi-scale approach: we present a spatially-dependent Cluster Dynamics (CD) model based on reaction-diffusion rate theory to describe the evolution in space and time of helium and its complexes. The use of CD modeling suffers from physical as well as computational challenges: physical challenges due to the large number of physical quantities (diffusion coefficients and binding energies) to determine for each chemical species, and the applied mathematics and computational challenges due to the large number of species that are modeled. The CD model is parameterized using MD simulation results, and a solver was developed to efficiently deal with the large system of non-linear partial differential equations describing the microstructure evolution. The solver uses an algorithm designed to exploit the local structure of the chemical reactions and makes use of parallel computing to both speed-up computations and reduce the memory requirements per processor. We compare the model results with results from molecular dynamics simulations of helium cluster formation and evolution below the surface, results from experiments performed using thermal desorption spectroscopy, and results from experiments under fusion relevant conditions.