The ASCI Challenge: Multi-Tflop Applications By 2002
Room N7
Chair: Gilbert Weigand
Department of Energy
Speakers: William Camp, Sandia National Laboratories; John Hopson, Los Alamos National Laboratory; David Nowak, Lawrence Livermore National Laboratory
This session will introduce the Department of Energy Accelerated Strategic Computing Initiative (ASCI), its goals and challenges. An overview of ASCI will be followed by presentations from each of the ASCI laboratories, LANL, LLNL and SNL, about the computational demands of the ASCI applications.
Two follow-on sessions will be held in the Allegheny Ballroom, Doubletree Hotel, at 1:30 and 3:30.
The ASCI Challenge: 100-Tflop Platforms By 2002
Allegheny Ballroom, Doubletree
Chair: Alex Larzelere
Department of Energy
Speakers: Vendor Representatives from each ASCI Platform (IBM, Intel, etc.); Robert Deri, Lawrence Livermore National Laboratory
This session will focus on the computational platforms that make up the ASCI program. A representative from each of the initial ASCI vendors will describe the special features of the ASCI Red and Blue platforms and their time tables for development. This will be followed by a description of the FY97 ASCI Engineering and Development initiative which will concentrate on technology research to break through barriers to attaining 100+ TeraFLOP/s performance by the end of 2002.
Supercomputing Applications In Medicine
Room N7
Moderator: John Gilbertson, The University of Pittsburgh
Speakers: Michael Becich, John Vries, The University of Pittsburgh Medical Center; Art Wetzel, Pittsburgh Supercomputing Center
Computing systems have become central to medical care and research, and their success has created a demand for systems with ever- increasing capabilities. Some applications are reaching computing power, bandwidth and storage requirements associated with supercomputers. The multimedia electronic patient record is such an application, especially at major medical centers where large scale image storage is becoming common. At the University of Pittsburgh Medical Center, a supercomputer-like machine, called MARS (Medical ARchival System), is now used for this purpose. In addition, current high-end workstations based on the Alpha and other leading-edge microprocessors now have the compute power and memory bandwidth of supercomputers 10 years ago. Assuming this trend continues, we should ask what medical applications the supercomputer community should be developing today that may run on desktop machines in the next decade.
The ASCI Challenge: Usable Environments For Terascale Computing
Allegheny Ballroom, Doubletree
Chair: Sean Headrick
Department of Energy
Speakers: Dona Crawford, Sandia National Laboratories; John Morrison, Los Alamos National Laboratory; Mark Seager, Lawrence Livermore National Laboratory
This session will start with an overview of the challenges associated with absorbing an increase in computing power by three orders of magnitude in less than two years. Challenge areas are high performance storage, portable applications development environments, distributed resource management and Tri-Lab Networking. The overview will be followed by a presentation on specific end-to-end solutions in the areas of visualization and interactive data manipulation to enhance scientific understanding. The session will conclude with a description of the ASCI program for Strategic Alliances intended to involve the academic research community in addressing the real problems faced.
Ten Years Of K-12 High Performance Computing And Communications: What Have We Learned? What Are We Building For The Future?
Room N7
Moderator: John Ziebarth, National Center for Supercomputing Applications
Programs such as Adventures in Supercomputing (AiS), the Alabama Supercomputing Program to Inspire computational Research in Education (ASPIRE), Earth Vision, the National Education Supercomputer Program (NESP), the PSC High School Initiative, SuperQuest and others have been offering in-depth training and experience with computational science tools and methods to teachers for a decade. This roundtable will offer a forum for those involved to discuss the successes and failures of these efforts and to communicate needs and goals for future programs.
DoD High Performance Computing Modernization Program
Allegheny Ballroom, Doubletree
Invited Speaker: Anita Jones, Director, Defense Research and Engineering
Moderator: Kay Howell, Director, DoD HPC Modernization Program
Special Guest: John Toole, Director, National Coordinating Office
Panelists: Presentors and discussants representing the three Program Initiatives (Centers, Networking, Sofware), the User-based Computational Technology Area (CTA) Groups and the Program's Academic Partners (Programming Environment & Training)
This panel presents various aspects of the Department of Defense (DoD) $2 billion HPC Modernization Program. The Modernization Program is establishing and managing the high-end commercial state-of-the-art computational infrastructure essential to maintaining and extending the United States' technological edge on the battlefield. These facilities currently include sixteen HPC centers, high speed network services connecting the centers and the highly dispersed user community, and major software initiatives in both scalable applications and programming environments and tools. However, the focal point of the program is the user community of over 4,200 active computational scientists and engineers employed at more than 100 laboratories, universities, companies or DoD facilities.
Following a forward-looking presentation by Dr. Jones, Kay Howell, Director of the DoD Modernization Program, will then lead a team of panelists in an interactive discussion of the program history, objectives, structure and strategy for addressing DoD needs in collaboration with the HPC community at large. The panel will highlight key projects, identify resources that will be available to DoD scientists and their collaborators and discuss the role of the academic partners in the program.
Four roundtable sessions in the afternoon will feature panels of leading computational scientists and engineers representing four of the ten computational technology areas: computational fluid dynamics, environmental quality modeling, computational structural mechanics and computational electromagnetics.
Computation And Competitiveness--High Performance Computing And Industrial Leadership
Room N7
Moderator: Peter R. Bridenbaugh, Alcoa
Panelists: Marvin Bloomquist, Mobil; Jack Thompson, Chrysler; Kenneth Neves, Boeing Information & Support Services
The effective use of knowledge lies at the heart of industrial leadership. This panel will discuss the role of high-performance computing in creating a competitive advantage in the aerospace, automotive, materials and energy industries. Panelists will also offer their view of the future for HPC and their requirements of the HPC community.
Center Directors Roundtable
Room N7
Moderator: Andy White, LANL, in coordination with Ken Kliewer, ORNL
This session will includethe following topics:
DOD Roundtable 1:
HPC For Computational Fluid Dynamics
Allegheny Ballroom, Doubletree
Moderators: Jay Boris, Naval Research Laboratory
2:15
DOD Roundtable 2:
HPC For Environmental Quality Modeling
Allegheny Ballroom
Moderator: Jeffery Holland, Army Corps of Engineers
Panelists: Joan Novak, EPA National Exposure Research Lab; Steve Yabusaki, Pacific Northwest National Lab
This panel will address the application of high performance computing to environmetal quality modeling. Particular projects include (a) the integration of multiple environmental quality modeling tools and HPC at the U.S. Army Engineering Corps Waterways Experiment Station, (b) research in subsurface modeling with an eye toward the geochemical modeling, as well as flow and transport, and (c) HPC research thrusts lead by the EPA in the air quality modeling area and in multi-media modeling.
DoD Roundtable 3:
HPC For Computational Structural Mechanics
Allegheny Ballroom, Doubletree
Moderator: Kent Kimsey, Army Research Laboratory
Panelists: Fouad Ahmad, National Center for Supercomputing Applications; Mark Emery, Navala Research Laboratory; Gene Hertel, Sandia National Laboratories; Raju Namburu, Army Corps of Engineers
This panel session discusses HPC aspects of computational structural mechanics. The panel discussion will focus on high-resolution, multi-dimensional modeling of materials and structures subjected to a broad range of loading conditions including static, dynamic and impulsive. Panel participants will address application-domain issues and computer-science issues, as well as the benefits and challenges of scalable computational structural mechanics application software.
4:15
DOD Roundtable 4:
HPC For Computational Electromagnetics
Moderator: Joseph Shang, AF Aeronautical Systems Center
Panelists: Louis Medgyesi-Mitchang, McDonnell Douglas Corp; Thomas Blalock, Missile and Space Intelligence Center; Lennart Johnsson, University of Houston
The computational electromagnetics panel will address the needs for modeling and simulation of electromagnetical phenomena by the defense industry and the theater missile defense program. An assessment of the current status and future trend of scalable parallel computing in computational electromagnetics will also be given. Finally, the approach and new opportunity of the DoD HPC Modernization Program Office's Common High Performance Computing Software Support Initiative will be presented and discussed.
Archival Storage Systems Experience For High Speed ComputingÑDiscussing The Features, Faults, Futures And Feasibility
Room N7
Moderator: Alan Powers, NASA Ames
Speakers: Alan Powers, John Lekashaman, NASA Ames; Terry Jones, Naval Oceanographic Office; Ellen Salmon, Douglas S. Carlson, Cornell Theory Center
Representatives of four sites, each running a different archiving solution, will form the panel. Each panel member will present information about their archive solution and usage statistics. They will discuss their perspective about the archive solution's features, faults and future enhancements. Each panel member will provide their software and hardware solutions and costs for three proposed archive requirements (5 TB, 125 TB, and 1000 TB). A lengthy question and answer session will follow the discussions.
Petaflops Computing
Room N7
Moderator: Paul Messina, Caltech
Petaflops computing has been explored through a series of federally sponsored community led forums. Although the feasibility and application of petaflops systems within the next twenty years are considered a distinct possibility, there are highly conflicting views concerning the means by which this is to be achieved. Such issues as architectural capabilities, system software, user interface, exotic technologies, economic business models in support of industry, and research and development paths are all considered controversial. Also, diverse perspectives are sometimes in direct conflict. For example, one approach dictates the use of commercial off-the-shelf components as the only practical approach to funding the necessary hardware and software technology. Another approach considers alternative structures of processing logic and memory on a single chip for more efficient computing. A third approach proposes to exploit exotic technologies in hybrid organizations, including superconducting logic, optical communications and storage, and semiconductor memories. At the same time, it is quite possible that the first petaflops computer will be a special-purpose device good for only a single class of applications but available in just a few years. Other approaches have been considered as well. The panel will present an array of such views and invite the audience to contribute their own ideas on possible paths to petaflops computing.
Scalable I/O Initiative Roundtable On Portable Programming Interfaces For Parallel Filesystems
Room N7
Moderator: Jim Zelenka, Carnegie Mellon University
The Scalable I/O (SIO) Initiative intends to release for public review a draft parallel filesystem API suite no later than six weeks before this conference. The goal of this suite is to define a common set of powerful interfaces for application, compiler and toolkit programmers to use for high-performance I/O on a variety of parallel machine architectures. In this roundtable, the SIO Initiative would like to collect, interpret and discuss the community's responses to, and suggestions for, the proposed API suite. Broad-based comments are sought, but we will be especially interested in the programming experiences of I/O-intensive applications, out-of-core compilers/toolkits and parallel file system implementations.
Impacts Of The Telecommunications Act Of 1996 On The HPCC Community
Allegheny Ballroom, Doubletree
Moderator: Dave Farber, The University of Pennsylvania
The Telecommunications Act of 1996 has made major changes in the rules of engagement of the key stakeholders in the telecommunications field. Key FCC rulings on interconnectionÑRules, Universal Service and Access ReformÑhave recently been or will soon be made. The panelists are drawn from the local, long distance, cable, metropolitan, cellular, FCC and internet communities. A brief summary by each panelist of the key impacts already seen will lead to a lively debate.