Vision
21st Century military superiority lies in the integration
of information technologies throughout the DOD global infrastructure
down to the level of the individual warfighter
Several fundamental changes are rapidly occurring in these
underlying information technologies that will radically enhance
the capability of the DOD to meet its mission
computing
single processor to parallel processors
knowledge space
user's hard disk to world wide web
mode of work
single user to collaborative virtual team
networking
low speed LAN to broad band national scale enterprise
human computer interface
text and image to immersive virtual environment
and digital multimedia
supercomputers
stand alone machines to distributed scalable
servers
science and engineering simulation
low resolution single physics, single scale models
to high fidelity multi-physics and multi-scale optimizations
How can DOD deploy a global-scale world-class IT infrastructure
to support R&D that can continually evolve to maintain state-of-the-art
capabilities?
Integrated but distributed information infrastructure must
be deployed
Production supercomputing capability needs to be brought in
rapidly
New generation of HPC software must be developed
to utilize new IT capabilities
Leveraging the state-of-the-art as it moves to
the state-of-practice
Continuous awareness education and training of
DOD community
DOD program components
DREN
MSRCs
DCs
CHSSI
PET
Why PET?
HPCMP goal: "Modernize the DOD R&D HPCC capacity
and capability to a level equal to or above the foremost academic
and industry research centers"-DDR&E
Synergy of academic and dod HPCC communities
technology transfer of academic innovation to DOD users
DOD problem solving drives academic innovation
Seeds for these technologies were sown by DOD thirty
years ago, the early prototyping was done in universities, and
the systems are now transitioning to industry. The PET program
provides a highly leveraged method for transferring the academic
experience base into the DOD mainstream.
Focus on enhancing user capabilities by effective use of new
generation of scalable computers
Create an Interlinked National-Scale User Community
Enable Collaborative Virtual Teams to attack Mission Critical
Problems
Objectives of this meeting
- Establish set of goals
- high-level goals
- level-2 objectives
- Methodology for identifying user and classifying targets
- Challenges
- Strategies for meeting goals
- infrastructure-related
- communication
- computation
- education
- culture-related
- process-related
- Recommendations
PET Program-wide Vision
- Meeting contractual obligations
- User community is aware of and has ready access to resources
/ methods enabling military successes
- High-impact point successes
- A specific user team accomplishes something viewed DOD-wide
as a success
- e.g., ability to prevent one military disaster
in future
- Basic research successes in key
DOD areas
- General-diffusion successes
- Develop or enable an application
that allows end-user group to solve significant problem
- Time frame for achieving many of these
is years, not months
Program-wide User Base
- Short-term (quick success) targets
- Defense Grand Challenge
- Peer-reviewed proposals
- Some similar to NSF Grand Challenges;
some very mission-specific
- Must use multiple MSRCs
- Must deliver within a year
- Other groups
- Projects interesting in and of themselves
- Projects likely to have widespread
future impact
- Medium-term targets
- Establish infrastructure and training
benefits
- CTAs
- Longer-term targets
- Spin-out benefits for general MSRC
communities
- Need education, not just training
or infrastructure
Impediments
- Infrastructure-related
- Serious gaps in computational infrastructure
- Culture-related
- No clear reward for users participating in educational activities
- Establish multi-university (joint) degree or
certification programs for DOD users
- Seed the community to help itself
- ìActuarial approachî is the one
that works best
Formalizing Program-wide Coordination
-
Andresís stuff describes overall purpose of
committee
-
Our piece does the specific objectives and mechanisms
for committee
- Establish academic PET executive committee
- Ken Kennedy, chair
- one (two?) rep for HBCU sites
- lead academic affiliates
- other major (in terms of responsibilities and funding) academic
sites
- (for formal votes, each lead academic affiliate will have
one vote)
Establish program-wide information base on user needs
and resources for addressing needs (e.g., software, algorithms,
architectures) and outcomes of PET activities
Observe and coordinate program as a whole, ensuring
that overall direction is consistent with objectives of HPCMP
and the specific MSRC contracts
- Move toward a common infrastructure to
support the S&T community
- Ensure appropriate links to national HPCC
community
- Identify structural impediments to progress and recommend
ways to overcome or eliminate them
- Establish program-wide efforts
- identify areas of overlap and commonality
- identify opportunities for new program-wide activities
or activities that should be phased out
- set up working groups to ensure integration and
coordination where appropriate
- Monitor program-wide efforts
- ensure balance among affiliates
and MSRCs
- arrange for resource pooling
- doument shared accomplishments
- publicize shared successes, giving credit to
all four MSRCs as well as end users
- Meet regularly with representatives of
the integrators, MSRCs, and CTAs
Program-wide Activities
- Ongoing activities
- Common Web document infrastructure (Baker/Coleman)
- Coordinated training (Southern/Turcotte)
- Electronic collaboration infrastructure (Ferguson/Turcotte)
- User profiling and analysis (Pancake/Blanchard)
- CTA-specific software repositories (Bernholdt/Grosh)
- How activities get adopted
- Approval sought on two coordinators (academic and government)
- Committee agrees that it is appropriate for program-wide involvement
Examples of Potential Program-wide Activities
- Scalable computing
- Metacomputing infrastructure
- CTA-specific software infrastructure
- Grid/geometry systems
- Discretization methods
- Solvers
- Virtual environments
- Data handling and analysis capabilities
- Multidisciplinary PSEs
- Enterprise-scale infrastructure
- Coordinated network connectivity
(hardware and software enabling)
- MSRC-wide file transfer and sharing
- Software interoperability
- MSRC-wide software configuration
management and improvement
- Education and training
- Web-based training infrastructure
(HBCUs as beta sites)
- Scalable education program
- Joint postdoc and internship programs
- Problem-oriented knowledge base
Specialization Breakdown