|
SC Global Event 2001 home | sponsorship | constellation sites | info for sites |
|
|
SC Global Technical Program Details Oil & Gas:
Problems, Research, and Tools in the HPC Field Bertrand Duquet - IFP These workshop sessions aim to exchange research challenges, solutions, experiences and knowledge related to high performance computing applied to the oil and gas sector. Namely the workshop will point out the problems faced by oil companies, their activities in the computational field and their strong relations with the HPC world. Moreover, the state-of-the-art solutions given by some hardware and software actors will be shown, related items as visualization will be investigated, research and computer science advances will be presented. The NSF Middleware
Initiative Enabling Scientific Process Across Global Grids Would you be interested in easier access to modeling software, supercomputing systems, large data sets - what about running real-time simulations over the global grid without a hitch? Join this discussion to learn how the NSF Middleware Initiative ("NMI") www.nsf-middleware.org will greatly enable industry, scientists and other grid computing users to seamlessly share applications, scientific and other specialized instruments, remote data sets and repositories around the world. User participants are currently being sought to participate in test trials! Access Grid
Middleware Requirements Moderator: Panelists: John Hurley, Clark Atlanta University Jon Swanson, inSORS Yoshio Tanaka, National Institute for Advanced Industrial Science and
Technology The challenges encountered in current middleware projects, such as authentication, authorization, resource access, resource discovery etc, have emerged and caused extensive attention while designing the next generation Access Grid (AG). We believe that the technologies having been implemented or to be applied in middleware projects will offer AG designers inspiration about the anatomy of the next generation AG. Participants in this panel will review the requirements and designing goals of the next generation of AG. Questions to be addressed include:
Grid-Net Access by
Everyone This BOF will explore the use of Grid-based modality-translation services to provide access by individuals in constrained environments or with disabilities. For example, a service which can describe visual information auditorially on request can be used for spontaneous participation by individuals who are blind or could allow someone to participate in an Access Grid session while driving. The ability to translate speech to text on the Access Grid could allow spontaneous participation by people who cannot hear speech well or are deaf, as well as making it easier for international participants. An Access Grid session that is text captioned can also be automatically indexed. Providing such services on-demand on the Grid could also be used by individuals with hearing impairment or deafness to to be able to participate in meetings or presentations (live or teleconference) at will and instantly call up speech-to-text translation that they need to participate. Finally, they can act as a safety net for Grid-based applications and services which are not easily accessible. The discussion will cover strategies for implementing such services as
well as implications for non-disability related applications in the context of Human Factors
and the Access Grid: Technology for Group Collaboration Panelists: Starkey Duncan - University of Chicago Elizabeth Churchill - FX Palo Alto Laboratory Angela Sasse - University of London The Access Grid (AG) provides a new frontier and opportunity for collaboration in scientific researches. This panel will provide some case studies on uses of Access Grid and related technology as well as live discussions on group interactions to address the human factors side of this new frontier. Any human factors study is either about the society shaping or socially shaped nature of technology. Hence our topics will be divided into these two aspects too. On the society shaping side, we would like to talk about how computer technology affects people's work (e.g. group-to-group communication, cross organization collaboration, new channels for scientific researches, etc.) On the socially shaped side, we would like to talk about how people's work practice (what tools they usually use, where they get helps, how they reach consensus and resolve conflict in their work, etc.), psychological (what is the motivation for people to work hard, to participate in teamwork, etc.) and biological attributes (verbal and non-verbal communication, personal space, etc), as well as their social organizational membership affect the effectiveness of the tools we try to build for them. Therefore, the following is a list of possible human factors topics we want to address in the panel: 1). Theories of human factors studies past and future; where we were
from, what we are doing, how will we progress; good and bad guidelines. Shrinking the Ponds History of the Trans-Atlantic and Trans-Pacific Metacomputing
Experiments from 1997-2001 Jodrell Bank Pulsar Search Code in the Metacomputing Experiments HLRS Applications in Molecular Dynamics and Fluid Dynamics Applications from Japan Metacomputing with Cactus This session will bring together representatives from major centres in Europe, USA and Asia who have been involved in experiments to exploit global metacomputing via trans-continental links. The global metacomputing started with a trans-Atlantic coupled application demonstrated at SC97, was extended to include trans-Pacific links at SC99 (winning an award in the HPC Games competition) and was extended at SC2000 to include all the above sites. There will be discussions of the results and experiences gained over the last few years, linking these with current initiatives in Global Grid Computing. The areas to be discussed are, message-passing libraries, intercontinental networking, scheduling meta-computing work across different machines, dynamic migration of applications. The scientific areas involved include, molecular dynamics, computational fluid mechanics, processing of radio astronomy data, and environmental prediction. This culmination of a project funded by the UK agency JISC and managed by UKERNA who provides the UK research network called SuperJanet. The purpose of this session will be to identify what applications are suited to run over inter-continental networks and what the potential scientific benefits could be. This session provides an opportunity to involve both computational and computer scientists in discussion leading edge problems that are potentially too large for any single machine. The method is via the coupling of large powerful supercomputers, thus the application fields are different to those exploited by the use of spare cycles on vast numbers of commodity PCs. Gordon Bell Finalists
Showcase Multi-teraflops Spin Dynamics Studies of the Magnetic Structure
of FeMn/Co Interfaces, An 8.61 Tflop/s Molecular Dynamics Simulation for NaCl with a
Special-Purpose Computer MDM Terascale spectral element dynamical core for atmospheric
general circulation models Achieving Extreme Resolution in Numerical Cosmology Using
Adaptive Mesh Refinement Resolving Primordial Star Formation A 11.55 Tflops simulation of black holes in a galactic center on
GRAPE-6 Supporting Efficient Execution in Heterogeneous Distributed
Computing Environments with Cactus and Globus Impact locating on aircraft structure by low-cost cluster 24.6
cents/Mflops ARSpace:
A Demonstration of Face-to-Face and Remote AR Collaboration The Human Interface Technology Laboratory at the University of Washington is a leader in the development of Augmented Reality interfaces for face-to-face and remote collaboration. Augmented Reality (AR) is a technology that enabled the overlay of three-dimensional virtual images on the real world so that both the real world and virtual imagery can be seen at the same time. The ARSpace interface is a generic platform for viewing and interacting with three-dimensional virtual models and supporting three-dimensional computer supported collaborative work (CSCW). As such it could be applied to a variety of application domains. In this session, an application in the area of geo-spatial data visualization, using environmental representations developed by the University of Washington's Center for Environmental Visualization (CEV), will be shown. Geospatial visualizations under development at CEV focus on the marine and coastal environments of the Northeast Pacific and the Pacific Northwest. Spatially registered representations of ecosystems and environmental processes are generated from scientific datasets from University of Washington Earth Science programs at the Big Beef Creek field station, multi-disciplinary PRISM and NEPTUNE research projects. For more information, see www.cev.washington.edu, www.depts.washington.edu/bbc, www.prism.washington.edu, www.neptune.washington,edu. Collaborative
Course in Parallel Scientific Computing Frank Williams, Director-Arctic Region Supercomputing Center, U. of
Alaska Fairbanks Frank Gilfeather, Director-Albuquerque High Performance Computing
Center, U. of New Mexico Guy Robinson, Arctic Region Supercomputing Center, U. of Alaska
Fairbanks Don Morton, Montana Rockies Center for Computational Science, U. of
Montana Jennifer Parham, Montana Rockies Center for Computational Science, U. of
Montana Tim Warburton, Albuquerque High Performance Computing Center, U. of New
Mexico Brian Smith, High Performance Computing, Education and Research, U.of
New Mexico Jeff Shuckra, High Performance Computing, Education and Research, U. of
New Mexico In this session, ARSC/UAF, Montana and New Mexico will present a description of their collaborative teaching strategies for a graduate level course, "Parallel Scientific Computation." The three universities are offering this course during the Fall 2001 semester; each site has a course instructor and students. Weekly Access Grid sessions between Alaska, Montana and New Mexico facilitate team teaching, as well as remote demonstrations, discussions and tutorials. Can
the Asia-Pacific Grid contribute to the Science and Technology in the Asia-Pacific Region? Moderator Dr. Paul Coddington, University of Adelaide, Australia Dr. Kazuhito Ohmaki, National Institute of Advanced Industrial
Science and Technology, Japan Dr. Sangsan Lee, Korea Institute of Science and Technology
Information, South Korea Dr. Fang-Pang Lin, National Center for High-performance Computing,
Taiwan Dr. Royol Chitradon, National Electronics and Computer Technology
Center, Thailand Dr. Mike Vildibill, San Diego Supercomputer Center, USA
This panel involves representatives from some of participating nations in the ApGrid and discuss how to contribute to activation of science and technology in Asia Pacific region. Workshop
Sessions on Grid Infrastructure Doing Metacomputing in UNICORE with Advanced Reservation Running jobs synchronously on different machines is a difficult task. Queueing systems that use common scheduling strategies like Backfilling or Gang-Scheduling can not guarantee the start time of metacomputing subjobs. As all subjobs need to be started at the same time, advanced reservation is the solution to the dilemma. We use PBS and our CCS (Computing Center Software) as local queuing systems which both support advanced reservation. Furthermore a mediator (called timer process) is needed which interacts with all queuing systems by querying the schedule and doing reservations. The timer process receives a metacomputing-AJO (abstract job object) from the gateway of the USite (UNICORE site, e.g. a Computing Center) which consists of sub-AJO's and an attribute specifying the start time of the whole metacomputing job. As the sub-AJO's are standard UNICORE-AJO's, they come with a complete job-description. This includes values like e.g. the needed resources (number of nodes, disk space, special software), the machine, or the path and name of the executable. The time attribute can have three different values: 1) An explicit starttime in a standard time format at which the timer process should start all subjobs. 2) The keyword NOW to depict that the timer process should try to start all subjobs at the time the metacomputing-AJO arrives. 3) The keyword ASAP (as soon as possible) specifies that the timer process starts all subjobs as soon as enough resources are available on all specified machines. When the timer process receives the metacomputing-job, all subjobs and their information are extracted. Then the timer process requests the current schedule from all machines respectively their queuing systems. With this information the timer process tries to find slots of free resources in the schedules so that all subjobs can be started synchronously at the given starttime. If a common slot is found, the timer process sends reservation requests to the local queuing systems. After the reservation is done and the reservation-ID is returned, the timer-process submits the sub-AJO's to the machines through the UNICORE hierarchy. In a first step metacomputing-AJO's are allowed only inside one USite, so that only inhouse-metacomputing is possible. Further development can lead to cross-site-metacomputing so that the timer process does not only contact local queuing systems, but also remote sites. The talk will describe the timer process (i.e. interfaces to local queuing systems, strategies for finding appropriate slots in the schedules, and creating the reservations) and its integration in the UNICORE architecture UNICORE-Globus Interoperation Getting the Best of Both Worlds Globus and UNICORE both establish a Grid infrastructure which gives users access to distributed resources. Both infrastructures offer a single sign on environment with X.509based security. Globus can be characterized as a tool kit that allows the development of Grid applications using a set of Globus services. UNICORE represents an integrated solution focusing on uniform access to distributed computing resources. UNICORE is developed by a consortium of German universities, research laboratories, and software companies. It is funded by the Ministry for Education and Research (BMBF). The goal of the work presented here which has been formed through a cooperation between Argonne National Laboratory and Forschungszentrum Juelich is to bridge between the two worlds combining the seamlessness of UNICORE with the flexibility of Globus. In the first step UNICORE will act as the client to access selected Globus resources, allowing for job submission, status queries, data staging and output retrieval. In addition, the objective was to realize the goal without changes to Globus and UNICORE. This is achieved through a spezified Target System Interface (TSI) whose role normally is to couple UNICORE and Batch systems. One of the significant subjects is the integration of the Globus security mechanisms in the UNICORE architecture. First of all, UNICORE and Globus use certificates issued by different Certificate Authorities with different policies. UNICORE guarantees the integrity of jobs by signing each part of a job with the user's certificate while Globus uses delegation. For the mapping of the two concepts the TSI has to obtain of a proxy certificate for the user to attach it to the Globus job it has generated from the UNICORE job.The TSI is capable of mapping UNICORE user requests, like job submission, status query,and retrieve output to the corresponding Globus user requests and it makes the Globus replies available to the UNICORE user. The presentation will give a detailed description of the established interface. In addition, solutions will be demonstrated with an application using UNICORE and Globus resources located in Jlich and at SC 2001. Broadening
the Grid - Increasing Minority Serving Institution Participation It is important that high performance networking and computing serve the interests of broader communities, especially underserved communities like minority communities in the US. Projects such as the EDUCAUSE/EOT-PACI Advanced Networking with Minority Serving Institutions (http//www.anmsi.org), SC’s Minority Serving Institution Participation Project have focused on including people from Minority Institutions in the design and utilization of the Grid. This BOF brings together participants in SC from minority serving institutions, groups interested in the 'digital divide' at the level of the Grid, as well as other interested parties to discuss how HPCC can help move underserved peoples and communities 'beyond borders'. The Primary Contact for this Activity is Stephenie McLean, the Access & Inclusion program coordinator at NCSA, who runs both the AN-MSI/EOT-PACI project and the SC initiative. Assisting will be Roscoe Giles, Boston University. Nodes to be included are not all determined yet. They include Boston, Access DC, Argonne, and hopefully Clark Atlanta, Dine, Albuquerque, and perhaps Houston. Online Digital
Property Management Panelists: Lindsay Moir - RightsMarket (rights management technology) Don Berkowitz – Digital Content Management Services (distribution
trials) Andrew Paskauskas - Sheridan College (digital content at educational
institutions) Tina Dalls - OnDisC Alliance Sheridan College will host a panel on the Management of Online Digital Property at Colleges and Universities, using the Access Grid to link panelists at several locations in the United States and Canada. A video describing the OnDisC - Online Distributed Content (www.ondisc.ca) - digital property rights management system will be a springboard for discussion of such issues as:
Education
Program - Distributed Mentoring Session During this period, Dr. George Francis, Professor of Mathematics from the University of Illinois at Urbana-Champaign, will be available to work with K-12 educators to advise them on tools, technologies, techniques and methodologies for integrating modeling and visualization into their mathematics courses. Dr. Francis will assist the teacher teams in defining and shaping the mathematics within their curriculum module so that the teams may integrate appropriate tools and approaches as they develop their projects. Dr. Francis will recommend materials and/or resources that would be relevant to the teacher team projects. Industry Spotlight
- New Applications Supporting BioTech "Internet2, Grid Computing, and Collaboration Research" High-performance computing has become a critical part of medical and biotechnology research, requiring much more computing power, collaboration, and expertise in these areas. Johnson & Johnson is bridging the gap between scientific research and advanced computing technologies to solve researchers toughest challenges. This presentation will provide an overview of how these new tools are giving J&J scientists the ability to virtually link together any facility, providing any user with unprecedented levels of productivity. There are endless uses including global virtual board meetings, distance education, scientific collaborations, and the linking of spaces like laboratories. By combining communications, global connectivity, we then begin to scratch the surface of the benefits that true collaboration can provide - to be productive and creative from anywhere, to anywhere in the world - at anytime! Global Grid
Forum Global Grid Forum (GGF), which began as a BOF at SC98, has emerged as an international organization with hundreds of participants from over 200 organizations and some 30 countries. GGF meets three times per year and currently includes over 20 working groups organized into 8 areas. After extensive discussions with, and advice from, leaders of the IETF and other standards bodies, the GGF has developed a standards process that is modeled after, and compatible with, the Internet Standards Process. This GGF BOF will provide a brief overview of GGF technical areas, structure, and standards processes. Education
Program - International Education and Training Forum (e-Learning) Advanced
Network and Application Research in China Education and research networks in mainland
China Dr. Xing Li will give a introduction about the history, the current status and the future plan of education and research networks in mainland China. Challenges of Computational & Collaborative
Environment in the Internet Prof. Wei Li will compare the Internet Computational Environment and Collaborative Environment, explain the technical challenges of these two, and give an introduction about the progress of a national key project named Massive Information Systems in Internet. Applications on HPNC (High-Performance
Networking and Computing) Environment Dr. Nai will introduce the recent research works and projects focusing on the advanced applications under a High-Performance Networking and Computing Environment, including some completed, such as "Scientific Database Applications on High-speed Network" and "Scientific Database Applications on HPC environment", and those in progress, such as "High Performance Streaming System" and "Scientific Data Grid". Workshop
Sessions on Grid Applications Density functional calculations on supercomputers The Nobel Prize for Chemistry was awarded in 1998 to Walter Kohn for the development of the density functional formalism. This method has been aworkhorse for over 30 years for solid state physicists who sought to interpret structural and spectroscopic measurements, but its general acceptance by chemists as a valuable method for calculating the properties of molecules came much later. These applications have been aided immeasurablyby the availability of supercomputers. I shall outline the basic formalism andthe nature of the numerical problems that it poses. One approach to their solution, developed originally by Car and Parrinello in 1985, is outlined anddemonstrated with the help of an interface developed recently within the framework of the UNICORE project. Development of a tool for performing numerical weather forecasts
via the Internet High Performance Computing centers in Europe as well as to operate aud support the EUROGRID software infrastructure, the EU-funded project "Application Testbed for European GRID computing" (EUROGRID) has been established. Within a grid of HPC centers of six different European countries three application oriented participants (University of Warsaw, Aerospatiale Matra CCR and Deutscher Wetterdienst) are applying their models of high complexity and resource requirements. The aim of the Deutscher Wetterdienst within the EUROGRID project (subproject Meteo-GRID) is to provide local weather prediction for arbitrary regions in the world via Internet and EUROGRID using the relocatable non-hydrostatic numerical weather prediction model LM (Local Model). This work involves by several steps. First a data set containing orographical height, land use type and soil texture is calculated from high resolution digital maps. The initial and boundary data for the Local Model will be interpolated from results of the Global Model GME, which are stored in an ORACLE data base. After providing the topographical, the initial and boundary data LM will calculate a 48-h forecast within a wallclock time of about 1 h. Using the EUROGRID software, the calculation of the topographical data set will be performed on the computers of DWD, whereas the synchoniously running interpolation and LM programs are applied on a HPC of one of the collaborating HPC centers. In the presentation a more detailed description of the steps towards an EUROGRID application of the LM will be shown. Some results of the LM are presented as an example of the numerous possibilities, which are given by numerical weather forecast. Characterisation and usage of a Virtual Intuitive Simulation
Testbed Franz Klimetzek, Ulrich Lang, et.al.DaimlerCrysler, Stuttgart and HLRS, High Performance Computing Center, University of StuttgartAbstractThe use of HPCN-technology enables industry to shorten development time by considerably reducing the computational time for numerical simulation. Current limitations for efficient usecome from insufficient through-put in the pre- and post-processing phase. The time toevaluate the huge amount of 3D data produced by numerical simulation and the time toredefine or to modify boundary conditions for derived simulations are the critical bottlenecks,that hinder the efficient use of simulation within the first phase of an industrial design, wherefast and frequent changes or modification of the components are inherent in the process. The intuitive simulation testbed Visit (Result of EC-Project VISiT,
ESPRIT No. 28247) basedon Virtual Reality technology, as a new man-machine interface is an
efficient environment fora small team of engineers and designers, that enables also the
designer to evaluate and interactwith the complex 3D data of simulations. This testbed
combines shortened pre- and post-processing time during successive design cycles with the
advantages of HPCN to improve the complete design and simulation process in the product
development. The efficiency of thedevelopment process was improved considerably, thus
leading to a strong reduction of costand especially of time-to-market for the European
industry. This improvement will be shownin different application fields: Solar Terrestrial Physics Sergei Maurits, Arctic Region Supercomputing Center Rob Baxter, University of Edinburgh Joanna Leng, University of Manchester An important scientific component of the above project is the work with the global solar-terrestrial physics community. They hve requirements for the coordination of data from a variety of sources, for example satellite, ground-based radar, magnetometer readings. Their data processing needs differ frm those of the Virtual Sky Survey and other similar projects, in that STP data is observed continually and unique events need to be processed along with long term trends on a variety of time scales. Solar-terrestrial physics presents challenges of linking data from instruments with computation, simulation and visualization, often under the constraint that this must be done in real-time to provide a space weather warning system. Our workshop will address this issue and will link this to longer term study of solar variability represented by historical records of sunspots and geomagnetic events such as aurorae. These workshop sessions will be of interest to both physicists and to grid middleware developers interested in addressing the issues of distributed working and collaboration under real time constraints. Grid Engine Project Compute farm clusters have made unprecedented computing resources available to developers and designers across a wide range of disciplines. From chip design to automobile design, and from animation to genome research, industry and academics share the need to make available large computing capability to individuals. Meet some of the people in the project and learn how Sun's new Grid Engine open source project can power a local distributed computing grid for your project in this technical review. BioCoRE -
Submits, Runs and Visualizes your Simulation from Afar BioCoRE is a next-generation web-based technology offering a seamless tool-oriented environment for structural biology. Within BioCoRE scientists can visualize information, share resources and interact with each other and with structural biology tools via a common infrastructure and across time zones and continents. During the demonstration we will use NAMD, our parallel molecular dynamics program and VMD, our molecular visualization program, and will perform an interactive molecular dynamics (IMD) simulation using NAMD and VMD in concert. With BioCoRE embedded tools we will feature the remote submission of a molecular dynamics simulation, the long-distance monitoring of the simulation's progress, the interactive steering of the simulation through a force feedback (haptic) device, and the visualization of the running simulation. Exploiting the high-speed NCSA Access Grid, BioCoRE participants at various grid sites will be able to join the multi-node-run session, to experience first-hand the haptic device, and share interactive graphics displays across distance. BioCoRE is developed by the Theoretical Biophysics group of the Beckman Institute for Advanced Science and Technology at the University of Illinois Urbana-Champaign. Challenges in Creating a National Collaboratory - The NEESGrid Example Moderator: Panelists: Steve Mahin - University of California, Berkeley Kim Mish - Lawrence Livermore National Laboratory Bob Reitherman - Consortium of Universities for Research
in Earthquake Engineering Bruce Kutter - UC Davis This panel will addresses the issues surrounding the creation of a national collaboratory, using NEESGrid as an example. The following aspects will be discussed:
Alliance
X-in-a-Box The NCSA Alliance In-a-Box initiative has recently released a set of standard software packages that lower the cost and the expertise needed to utilize four new technologies commodity computational cluster software, Grid toolkits, Access Grid collaboration systems, and scalable tiled displays. These packages create a new level of interoperability to support the needs of the national research community. At this BOF, potential users can learn about the latest status of these packages and query developers about the capabilities and future plans. Moderator: Panelists: Fabrizio Gagliardi - CERN Alex Szalay - Johns Hopkins University Yoshiyuki Watase - KEK Laboratory, Japan This panel will address the usage of data grids to address the unique challenges posed by data intensives sciences. Developing
an Australian Grid for National and International Cooperation
|
| ||||