Our goal is to build a community interested in developing Urgent Computing applications, improved infrastructure and middleware, and community policies. Together we can build a roadmap to improve support for applications
URGENT COMPUTING 2007
In a growing number of problem domains urgent access to large-scale computation is required. In some cases, key decisions concerning life-threatening situations must be made quickly with the aid of modeling and simulation. In others, unique scientific opportunities may be lost without immediate access to supercomputers. For urgent computations, late results are useless results. Computer models must be run and the data analyzed while their predictions can still be applied or the opportunity is available. These on-demand large-scale computations can't wait endlessly in a job queue for supercomputer resources to become available. Neither can the community keep multimillion-dollar infrastructures idle until required by urgent computation. We must leverage our existing investments in high-end computing and organize policies, tools, applications, workflows, and infrastructures to support urgent computing.
- The workshop will bring together four communities:
- Application scientists from a wide range of problem domains: earthquake, wildfire, severe weather, transportation, epidemiology, economics, and critical infrastructure.
- HPC Resource providers and Grids such as TeraGrid, DOE, universities, and NASA.
- Computer scientists and middleware providers.
- Policy and resource allocation experts. We anticipate approximately 40 attendees at the workshop.
April 25-26, 2007 (arrive the evening of April 24th, depart shortly after noon the 26th).
Argonne National Laboratory (ANL).
The workshop will take place in Building 221, Room A216.
Attendence is via invitation and our meeting space is limited. Please contact
Pete Beckman (beckman.at.mcs|anl|gov) if you have questions or would like
to participate. Registration is free. A visitor badge is required.
All participants must contact Lori Swift (swift.at.mcs|anl|gov) to get visitor passes to ANL. Lori will also help organize accommodations at ANL or a nearby hotel.
The agenda is still being developed, but will be roughly as follows:
8:00 - 8:30: Introduction [pdf] : Participant introductions, overview of workshop goals, introduction to urgent computing topics for workshop
8:30 - 10:10: Application Requirements: What are realistic scenarios for Urgent Computing applications? What computational requirements do they have, including real-time deadlines,
computational power, data movement, and real-time analysis and knowledge extraction?
- EpiSims & Simfrastructure for Modeling Pandemics [pdf]: Keith Bisset, Virginia Bioinformatics Institute
- LEAD and Severe Weather Modeling [pdf]: Dennis Gannon, Indiana University
- Wildfire Modeling [pdf]: Craig Douglas, University of Kentucky / University of Yale
- SURA Coastal Ocean Observing and Prediction (SCOOP) [pdf]: Gabrielle Allen, CCT/LSU.
10:10 - 10:30: Break
10:30 - 11:45: Application Requirements
- Drivers of URGENT-Computing Environments - Dynamic Data Driven Application Systems (DDDAS) [pdf]: Frederica Darema, NSF
- Earthquake and Geophysics [pdf]: Phil Maechling, Southern California Earthquake Center, USC
- Issues in Theater Deployment of a New Large-Scale System [pdf]: Alok Chaturvedi, Purdue
- NASA Space Shuttle Modeling [pdf]: Benjamin Kirk, NASA.
11:45 - 12:45: Lunch
12:45 - 2:50: Infrastructure: What infrastructure can be used today, and what must be developed to better support urgent computing? Can emerging checkpoint/restart technologies help yet? Can we accurately predict if deadlines can be met? How will we choose resources?
- SPRUCE Urgent Computing Environment [pdf]: Pete Beckman, ANL/UChicago.
- Batch Queue Prediction Strategies [pdf]: Rich Wolski, UCSB.
- Rapid Virtual Private Cluster Deployment [pdf]: Satoshi Matsuoka, Tokyo Institute of Technology.
- Checkpoint / Restart Technology [pdf]: Paul Hargrove, Lawrence Berkeley National Lab
- Job Schedulers, Resource Management, Reservations, and Preemption [pdf]: Bill Nitzberg, Altair Engineering.
2:50 - 3:10: Break
3:10 - 5:15: Demos, Resources, Policies, and Experiences: How does the deployed software work now? Can we support urgent computing on existing platforms? What resources are available for this new model? What is feasible today, and what can we look forward to tomorrow? What policies might we adopt to help grow the community? How can stakeholders support urgent computing?
- Demos of some existing systems [pdf] : SPRUCE and LEAD Spring Urgent runs.
- Support for urgent computing at SDSC [pdf]: Nancy Wilkins-Diehr, SDSC.
- The Virginia Tech System X [pdf]: Geoff Zelenka, Virginia Tech.
- The Elastic Computing Cloud at Amazon.com [pdf]: Martin Buhr, Amazon.com
- Experiences with Preemption [pdf]: Mike Papka, ANL/UChicago.
5:00: Adjourn: Head to local restaurants for further discussion.
8:30 - 11:00: Moving Forward Plans: Where do we go from here? How can we improve support for urgent computing, and what is our roadmap for success. What strawman policies could we propose? What application teams are ready for testing? What are we missing? How can companies support our work?
Panel: Roadmap Ideas for Policy, Tech, and Allocation:
Charlie Catlett, ANL/UChicago
Dennis Gannon, Indiana University
Anke Kamrath, SDSC
Satoshi Matsuoka, TITECH
John Towns, NCSA
Nancy Wilkins-Diehr, SDSC
Pete Beckman, Kelvin Droegemeier, Dennis Gannon, Jan Mandel, Nancy Wilkins-Diehr
Pete Beckman (beckman.at.mcs|anl|gov)