1. Computer and Information Science and Engineering Robert R. Borchers Division Director Advanced Computational Infrastructure and Research ORAP, March 21, 2000 The Information Technology R&D Initiative

2. 2 Overview ITR agenda maps PITAC into ITR and beyond Fundamental, long-term thematic IT research Advanced computing for science, engineering, and the Nation Ethical, social, and economic implications of Information Revolution Education and training of America’s IT workforce Five teraflop computing capability for S&E research FY 2000 Current Plan Information technology research - $90 million Terascale acquisition - $36 million

3. 3 ITR Research Agenda Increase fundamental, long-term, high risk basic research Software – security, reliability, scalability, ease of reuse Human-Computer Interface and Information Management High-End Computing Scalable Information Infrastructure Socioeconomic, Ethical, Legal & Workforce Implications

4. 4 A Vision of the World in 2010 2010 Social, Behavioral, Economic Sciences Information Sciences Life Sciences Physical Sciences, Engineering 2000 Social, Behavioral, Economic Sciences Life Sciences Physical Sciences, Engineering Information Sciences IT 2 Research

5. 5 Software Goals Increase productivity Reduce fragility Strengthen security Manifest adaptiveness and flexibility Methods Seek revolutionary ideas and validate on realistic problems Increase science base underlying software development Bring greater engineering discipline to development process Incorporate self-stability to achieve fault-tolerance, load- balancing, etc.

6. 6 High-End Computing Pyramid “Familiar problems are made more acute by greater scale, diversity, and performance requirements” – Herb Schorr

7. 7 High-End Computing Research Goals Develop enabling computational science technologies for all disciplines Advance the fundamental high-end computing state of the art Methods Encourage exploration of new ideas in scalable software and scientific visualization Launch a major program for scientific applications Establish cross-directorate activity

8. 8 Scalable Information Infrastructure: Ubiquitous Connectivity Satellite Global Suburban & rural Urban Pico-Cell Micro-Cell Macro-Cell In-building Pico-Cell Adapted from Tim Hewitt, “UMTS Overview,” TIA inf. Session, ITU Comf., Mpls, MN, Oct. 17-18, 1998

9. 9 Socioeconomic, Ethical, Legal & Workforce Implications Economic and social implications of IT Interaction between social and computer scientists Workforce Understanding the pipeline Technology for learning Innovation in education Mentoring

10. 10 The Presidential Information Technology Advisory Committee (PITAC) recommended:  Major new investments in long-term computing research, including high-end computation, and  Providing a terascale computing system to support future computational and computer science research. NSF’s implementation of a terascale system, following a full and open competition, would augment the resources available to the Science and Engineering computation community through the Partnerships for Advanced Computational Infrastructure (PACI) program. High End Computing in IT 2

11. 11 Presidential Advisory Commission on Information Technology High-end needs highlighted by PITAC –Long-term research in innovative computing technologies & software –Funding leading-edge computing facilities for research High-end recommendations by PITAC –Major new investments in long-term research enabling high-end computation –Terascale equipment acquisition supporting current computational and computer science research –Integrated research and development plan to create balanced petaflop systems by 2010 Addressing these will require coordinated efforts with other IT 2 thrusts

12. 12 HPCC Workshops Background NSF hosted 3 workshops in 1998 to identify HPC challenges Science workshop findings –Important applications need large scale computing, large memories, high bandwidth networking and I/O Algorithms workshop findings –Algorithm development requires teams with representation of applications and computer science –Algorithms must demonstrate usefulness for applications on BIG systems Software workshop findings –HPCC potential of large parallel systems can only be partially fulfilled due to limited programmability, limited understanding of end to end behavior, inadequate data management, … –Software research is necessary in all of these areas

13. 850 Projects in 280 Universities Universities With Projects Using the PACI Supercomputers

14. 14 Computational Science Terascale Examples Weather and Climate Prediction Protein Folding Cosmology Chemical Kinetics Gravitational Waveforms Seismology Combustion Systems Large Scale Economic Systems Analyses Materials Sciences

15. Capability Computing - Numerical Prediction of Thunderstorms Same Computer Forecast Without NEXRAD Radar Data Moore, OK Tornadic Storm Missing! Nat’l Weather Service Computer Forecast (unable to representindividualthunderstorms) Moore, OK Tornadic Storm 2-Hour ARPS Computer Forecast Down to the Scale of Counties -- Run on the 256-Node Origin 2000 at NCSA Central Oklahoma, May 3 1999, 7:00 pm CDT 7 pm NEXRAD Radar Observations Moore, OK Tornadic Storm

16. Five years ago, this sort of prediction was deemed theoretically impossible. Now it’s being done in real time! With terascale systems it will be routine. NSF Science and Technology Center for Analysis and Prediction of Storms University of Oklahoma

17. 17 Review and Selection Criteria The normal NSF review criteria will be used, plus examination of: –Delivery schedules, costs, and system performance –Evidence that the institution can mount and sustain the system –Construction and operations budgets, and sources of funds –Quality and credibility of the following plans: Physical Facility, Computer Acquisition and System Integration, Staffing & Management, Computing Resource Management, Network Connectivity, and Education.

18. 18 Goals To use information technology effectively and efficiently To make computers do what we want To find out what we wish to know Research Themes Active Knowledge Computer Human Boundary Human Support and Augmentation Human-Computer Interaction & Information Management

19. 19 Scalability Heterogeneity Self-management Adaptability Security/privacy Reliability Dynamic Networks Scalable Information Infrastructure: Central Research Issues

20. 20 ITR Multidisciplinarity Multidisciplinary research stressed Encourages projects simultaneously contributing to IT and other disciplines Cross-directorate collaboration and involvement Extensive consultation in planning process Multi-directorate ITR Coordinating Committee Advises CISE AD NSF-wide contact point Responsible for preproposal and full proposal reviews

21. 21 Status of ITR Research Competition Solicitation released Sept. 28, 1999 Project Budgets over $500K Letters of intent due Nov. 15, 1999 Preproposals due Jan. 5, 2000 (Panel review early Feb. 2000) Full proposals due Apr. 17, 2000 (Panel review mid May 2000) Project Budgets under $500K Letters of intent due Jan. 5, 2000 Full proposals due Feb. 14, 2000 (Panel review mid Mar. 2000)

22. 22 ITR Management Impact Assessment Annual awardee progress and final reports Midterm external reviews, site visits, and/or PI meetings for research project awards with budgets above $500K Internal tracking of all awards Reporting to external groups, e.g., PITAC, OSTP, OMB, and interagency ITR working group External Program Evaluation planned in FY2003

23. 23 Challenges and Opportunities Involving teaching faculty Role of research institutes Support for research/engineering staff Changing Federal IT support picture, e.g., Possible change in DARPA research emphases Possible increase NIH participation No FY 2000 Terascale hardware funding for DOE

24. 24 Challenges/Opportunities Forging collaborative interdisciplinary environments, e.g., IT and Biology in bioinformatics IT and Physics in quantum computing IT and Chemistry in molecular switching Computer Science and Computational Sciences...

25. 25 ITR Management Review Process CISE and other directorates participate, as relevant Multi-level Pre-proposal, full-proposal, site-visit (as necessary) Final award decision by AD/CISE in consultation with other ADs Awards 30% of program funding in ~ $2M-$4M/year awards 40% of program funding in ~$1M/year awards 30% of program funding in ~$150K/year awards