Computing Resources Overview
ASC Business Model
Terascale Simulation Facility
National Security Applications Co-Design Project
The NNSA labs and ASC program are defining a coordinated co-design strategy that focuses on the unique needs of the ASC program while leveraging the work of the ASCR centers. ASC is a mission-driven organization with applications currently in use that are of importance to run at exascale in support of stockpile stewardship, namely the Engineering and Physics Integrated Codes (EPICs) and their supporting codes and libraries. As such, the ASC project will focus on these established applications as the drivers, and participate in co-design largely through proxy applications.
National security EPICs uniquely need to be in service for many years, often a decade or more, because of their complexity and long development time. Many codes now reaching maturity in the design community were started 15+ years ago with the beginning of the ASCI program. Likewise, the codes must model a variety of physical configurations and allow the end user flexibility in their operation, sometimes through the use of code steering. Because each NNSA laboratory has their own set of applications and diversity of approaches, each is proposed to develop an autonomous co-design project, with strong coordination amongst themselves as well as the ASCR centers, particularly on areas of crosscutting research.
Focus and Goals
ASC is developing a tri-lab co-design project that focuses on the following key needs from the integrated codes mission. The National Security Applications (NSApp) Co-design Project will be realization of this concept with the following focus and goals:
- Next generation platforms delivered by the ASC Program must allow effective exploitation by ASC EPICs, not just science codes, to meet mission requirements.
- The emphasis on physics coupling in the ASC integrated codes presents an additional challenge that will likely be addressed only by the NSApp CDP.
- Large codes often stress the software stack much more than smaller codes, whether it's tools (debuggers, hardware simulators, etc.), compilers, or operating systems.
- Due to critical mission requirements, at-scale resiliency of the entire system – consisting of the hardware, the software stack, and the application codes – must be addressed so that simulations of interest can be completed when needed.
- Because of the size of these ASC EPICs (often > 1M lines), and the necessity to maintain the same code base for multiple platforms, code reuse and performance portability is a high priority. At first we expect each laboratory to set its own direction in the effective use of proprietary, non-standard, or vendor-specific methods. Eventually a standard programming model should emerge with the goal of making hardware-specific coding transparent to all application programmers.
NSApp CDP will use representative proxy applications as the means to interact with our vendor partner(s) during the co-design process. Proxy applications can be grouped into categories of increasing sophistication and fidelity in relation to the actual applications (or packages) used in integrated codes, ranging from small kernels to large integrated applications more closely representative of the real applications.
The increased emphasis on physics coupling in the ASC integrated codes presents an additional challenge that will be addressed by the NSApp CDP. When a synergy is not possible due to a unique characteristic of NNSA's codes, NSApp CDP will pursue co-design research independently with the appropriate vendor partner, with the goal of sharing generalized results (that is, those that do not compromise vendor intellectual property) with the DOE co-design community whenever possible.
Areas of Investigation
Physics and numerical algorithms that the NSApp CDP will investigate (and are candidates for algorithmic collaboration with ASCR CDCs) include:
- Deterministic transport
- Monte Carlo transport
- Diffusion processes
- Coupled hypersonic aero-thermal
- Structural mechanics, shock and vibration, and failure models
- Thermal effects and thermal-structural interactions
- Linear and non-linear solvers
- Subscale models
NSApp is also interested in exploring co-design in cross-cutting areas such as compiler development, system software (e.g., I/O, operating systems), and resiliency. We anticipate that much of this work will be collaborative with the ASCR CDCs.
Paul Adamson, ASC Integrated Codes Program Manager
National Security Applications Co-Design: A Framework for an ASC Tri-Lab Project [PDF]
Scientific Discovery through Advanced Computing (SciDAC): Co-Design
ExMatEx: Exascale Co-Design Center for Materials in Extreme Environments
CESAR: Center for Exascale Simulation of Advanced Reactors
Multi-disciplinary Research at Work: Combustion Exascale Co-Design Center Established