280 likes | 771 Views
Forces Modeling and Simulation Potential Areas for Future Collaboration. Dan M. Davis ddavis@isi.edu Information Sciences Institute University of Southern California. Central Idea.
E N D
Forces Modeling and SimulationPotential Areas for Future Collaboration Dan M. Davis ddavis@isi.edu Information Sciences Institute University of Southern California
Central Idea The Maui High Performance Computing Center and the Information Sciences Institute have several common goals within the general area of computational science and they have complementary capabilities for achieving those common goals. DoD Forces Modeling and Simulation (FMS) is a paradigm example of a research discipline that would and should benefit significantly from a collaboration between ISI and MHPCC.
Topics to be Covered • A Short History of FMS • A Quick Look at Previous Collaboration • A Review of Current Activities • An Outline of Future Work Planned • An Analysis of Potential Users/Funders • A Vision for the Future
FMS History • Battlefield analogs - as old as battles • Chess • Olympics • Goals from the beginning • Training • Analysis • Intentions • Assure victory • Reduce losses
Early Battlefield Modeling • The Board Simulations • Popular in last two centuries • Origin of the term “pucker” • Examples • US Navy - Pearl • Japanese Navy - Midway
The Computer Arises • ASW Simulator • JANUS • DARPA Simulation Networking (SIMNET) • The Joint Training Confederation (JTC) • Modular Semi-Automated Forces (ModSAF) • Synthetic Theater of War (STOW)
FMS Family Tree A Mandatory Graphic Albert UKCATT SAF StandAlone BothFOR CCTT SAF Linux ODIN SAF OTB SIMNET SAF ModSAF VERTS DI SAF JSAF
Previous Collaborations • Synthetic Forces Express • Major collaboration in late 90’s • MHPCC primary source of cycles for big runs • Project Albert • Some cross-fertilization • Different level of models • JESPP • Ongoing collaboration • Model for future work
SF Express • DARPA and The STOW Constraints • SF Express 1995-1998 • Scalable Implementation of ModSAF • Introduction of Software Routers • Globus as a facilitator • Achievements • March 1998 – 108K Operational Entities • 9 Centers (Maui to Maryland, Tenn. to Texas)
JESPP • Joint Experimentation on Scalable Parallel Processors • Workshops run by JFCOM/IDA • Spring 2002 • Cray Henry remarks • Funding via JFCOM – multi-year • Continuation of scalability via routers • HPCMP Cluster Symposium Visibility
Current Activities • JESPP • Cluster utilization • Interactive Computing
JESPP • Running well • Series of “Experiments” • Emphasis on urban warfare • Interest in asymmetrical warfare • One Week/Month => Two Weeks/Month • Goal: “develop future joint warfighting concepts” • Adm Edmund Giambastiani • Current Commander JFCOM • Director Joint Experimentation – MajGen Dubik, USA
Cluster Utilization • Clusters optimal for some work • Integer vice floating point • Inter-node connectivity • SAFs and other Simulations • Preponderance of integer math • Designed to tolerate latencies • Ease of implementation
Interactive Computing • HPCMP focus on batch • HPC community and interactive computing • Communication issues for MHPCC • 6K miles from “Right Coast” • Latency • Speed of light • Multiple physical router latencies • Bandwidth • Reliability
Future Work Planned • Work with JFCOM • JESPP • Routers • Logging and storage • Data mining and analysis • Visualization • Work with PEO STRI • One SAF • Implementing software routers
Potential Users and Funders • JFCOM • DARPA • DHS • ARDA • HPCMP • Services • PEO STRI (US Army Orlando) • SIMAF (USAF at WPAFB) • ACETEF (USN NAVAIR at Pax River) • MCWL (USMC at Quantico)
JFCOM • Funding through next several years • Giambastiani may be key • Need to set new goals • Help develop approach to future • Need visibility at higher level • Provides good vehicle for approaching HPCMP and other due to “Jointness”
DARPA • The Death of Simulation • New Leadership • The Market for Change • Talking to People Inside • May be good Target for: • Quick/small money • Major new initiative • Targets of opportunity
ARDA • NSA’s DARPA • Analysts’ trainers • Battlefield data fusion
HPCMP • PET program • Platform for other simulations • Need for FMS advocate • Funding
Services • PEO STRI (US Army Orlando) • SIMAF (USAF at WPAFB) • ACETEF (USN NAVAIR at Pax River) • MCWL (USMC at Quantico) • Joint - J7 and DMSO • Army - FCS • Navy – SPAWAR SSC Code 44 • Air Force – ESC Code CXE • Marine Corps – PM TRASYS
A Five-Year Vision • Secure the JFCOM Beachhead • Press ahead with HPCMP/PEO STRI • Submit CHSSI Center of Excellence Proposal • Fill FMS power vacuum • Make Maui “center” of FMS
Securing the JFCOM Relationship • Operational Mandate • ~3 FTE’s to support ops • Funding for the foreseeable future • Logging and Analysis • Div 2 personnel • Under IDA tutelage • Should expand to cover some needs • Run SLAMEM on Cluster • Implement on other platforms
Center of Excellence HPCMP CHSSI • Start and FY05 CHSSI program to implement a large-scale entity HPC FMS code set • Leveraging existing codes / projects • OOS – Simulation engine / Behaviors • JESPP (SFExpress) – Data Distribution • DCEE – Communication / Interoperability standards • MPARS / DCAPES – Exercise Setup / Review • Preliminary studies should commence immediately to identify, prioritize and quantify the work that needs to be done
Supporting Arguments • FMS is both an HPCMP S&T and Operational CTA • DoD is taking a larger operational focus • funding and relevance • HPCMP gains champions in the “mainstream” DoD • FMS users gain a viable new tool and mode of operation • As an S&T project, it is not bound by excessive conflicting requirements • System implementation of may one of the “hard problems” in parallel and distributed computing
FMS Today • DMSO – not as powerful • JSIMs – Memo terminating after FY 03 • DARPA – Reawakening some interest • ISI/MHPCC can become major force if • Successful • Visible • Accessible
Maui - Center of FMS Universe • Has • Compute power • Reputation • Mandate (can do classified ops) • Room to expand • Support of community • Relationship with teammates with virtually unlimited research resources
Conclusion MHPCC and ISI can become both the drivers of and the leaders in US Forces Modeling and Simulation. To not do so not only misses an incredible set of opportunities, but may border on dereliction.