Berkeley Lab to Showcase HPC and Networking Leadership in Talks, Demos at SC05

Computing and networking experts from the U.S. Department of Energy’s Lawrence Berkeley National Laboratory will share their leadership expertise via talks, technical papers and demonstrations at the SC05 conference to be held Nov. 12-18 in Seattle.

Sidney Fernbach Award
John Bell, a senior staff mathematician and head of the Center for Computational Sciences and Engineering at LBNL, has been named as the recipient of the 2005 Sidney Fernbach Award. The Fernbach Award is given by the IEEE Computer Society for an outstanding contribution in the application of high performance computers using innovative approaches. The award will be presented at the SC05 conference in Seattle. In conjunction with the award, Bell will give a special talk in a session starting at 8:30 a.m. Wednesday, Nov. 16.

Technical Program Presentations
Berkeley Lab is also well represented in the SC05 technical program, with LBNL staff authoring or co-authoring the following five technical papers:

• “Analyzing Ultra-Scale Application Communication Requirements for a Reconfigurable Hybrid Interconnect,” by John Shalf, Shoaib Kamil and Leonid Oliker of the Computational Research Division (CRD) and David Skinner of NERSC.
• “Leading Computational Methods on Scalar and Vector HEC Platforms,” by Leonid Oliker, Michael Wehner and Andrew Canning of CRD, Jonathan Carter of NERSC and others from Princeton Plasma Physics Laboratory, Lawrence Livermore National Laboratory, NEC, Oak Ridge National Laboratory and JAMSTEC.
• “Apex-Map: A Global Data Access Benchmark to Analyze HPC Systems and Parallel Programming Paradigms,” by Erich Strohmaier and Hongzhang Shan of CRD.
• Kathy Yelick, leader of CRD’s Future Technologies Group, is an author of “Making Sequential Consistency Practical in Titanium” along with her UC Berkeley students  Amir Ashraf Kamil and Jimmy Zhigang Su. Their paper has been nominated for the Best Student Paper Award.
• Additionally, Erich Strohmaier is a coauthor with UC San Diego researchers of a paper on “Quantifying Locality in the Memory Access Patterns of HPC Applications.”
• Mike Welcome of the Future Technologies Group, along with scientists from LLNL, coauthored a paper on “Tera-Scalable Algorithms for Variable-Density Elliptic Hydrodynamics with Spectral Accuracy.”

 On Sunday, Nov. 13, an LBNL team will present a tutorial on “Keeping Ahead of the Bad Guys: High Performance Computing Protection.” Presented by LBNL’s Stephen Lau, Scott Campbell and Bill Kramer, this session will focus on what can sites do to combat attackers while maintaining an open scientific environment.

Booth Talks and Demonstrations
As part of the series of presentations in Berkeley Lab’s booth (1828), representatives from DOE’s three INCITE (Innovative and Novel Computational Impact on Theory and Experiment) projects will discuss their latest work into how stars and solar systems form, how to make combustion more efficient and how proteins express genetic information.

Other talks in the LBNL booth will focus on what’s new at Berkeley Lab and NERSC, high performance computing benchmarking and performance analysis, large-scale calculations for nanostructures, cyber security, 10 gigabit testing on ESnet, query-driven visualization and the ACTS Collection of HPC tools. In addition to the three INCITE projects which compute at NERSC, another group of users will discuss “Core Collapse Supernovae: Trying to Explode Stars at NERSC.” LBNL is also hosting a talk by two Louisiana State University researchers whose planned booth fell victim to Hurricane Katrina. A complete schedule can be found at <http://www.lbl.gov/CS/Archive/SC05_Speaker_Sched.pdf>.

Berkeley Lab will also present demonstrations of a number of tools and techniques developed to advance scientific computing and networking. Booth demonstrations will include the following:

• Large-Scale File Replication using DataMover Technology
• Integrated Performance Monitoring of HPC Workloads
• Integrating System-Wide Global Parallel File System with HPSS
• The Warewulf Cluster Toolkit
• The Advanced CompuTational Software (ACTS) Collection
• Serial, Parallel and Distributed Checkpoint/Restart for Linux
• Enabling Reliable, Secure Group Communication
• ViCE: Visual Component Editor
• Weather via Open InfiniBand
• Drosophilia Gene Expression Visualization
• Interactive Single Nucleotide Polymorphism Visualization
• SciDAC Visualization: Accelerator Modeling Pilot Program
• Visualization of NERSC Computational Science in Astrophysics, Combustion and Structural Biology

SC05 Challenges
Berkeley Lab scientists will be participating in all three challenges at SC05: HPC Analytics, HPC Bandwidth and StorCloud, which push the boundaries of high performance analytics, storage and networking applications.

A team led by Will Baird is vying for the Tri-Challenge crown by compiling the highest combined score of the three challenges. The entry is called TRI Data Storm and will be demoed in the LBNL booth.

 Another Berkeley Lab team led by Kurt Stockinger is competing in the HPC Analytics with an entry called "Network Traffic Analysis with Query Driven Visualization.” This entry will also be shown in the LBNL booth.

Conference Organization
Finally, Berkeley Lab’s involvement in this year’s conference literally starts at the top — NERSC General Manager Bill Kramer is the general conference chair for SC05, and his executive assistant, Zaida McCunney, has served as vice chair of the conference committee. Other LBNL staff members have been active in the networking, communications and technical program aspects of the conference.