NASA CFD Vision 2030: The Power of Exascale Computing
1. CFD Vision 2030: The Power of Exascale Computing
NASA FUN3D Team is realizing substantial performance benefits
Challenge
The emerging landscape of exascale supercomputers
offers unprecedented computational power, but
porting applications to use these systems effectively is
a daunting proposition for software developers.
The CFD Vision 2030 Study found HPC to be a key enabling technology
Expected Impacts
- Tackle previously intractable grand challenge problems
across all NASA missions
- Profound impact on engineering-class applications
across the aerospace industry
- Substantial reductions in power consumption and
footprint in the data center
Collaborations
- NASA ARMD, Transformative Aeronautics Concepts
Program, Transformational Tools and Technologies
- DOE Argonne and Oak Ridge National Laboratories
- Department of Defense
- Georgia Tech and Old Dominion University
- AMD, HPE/Cray, Intel, NVIDIA
Solution
- The FUN3D team is collaborating with numerous organizations to
realize a next-generation High Performance Computing (HPC)
capability for NASA’s Computational Fluid Dynamics (CFD) needs
- Vastly more powerful computing enables scientists and
engineers to rapidly simulate environments that are infeasible or
prohibitively expensive for physical testing
- With FUN3D’s broad distribution to hundreds of groups across
industry, academia, and other government agencies, benefits are
widely available to the U.S. aerospace community
Results
- Performing high-fidelity simulations on emerging architectures at
leadership-class scales, reducing computational times by orders of
magnitude over conventional architectures in capacity environments
- External users report a 10x reduction in power consumption while
saving precious floor space in data centers
- Engineering-class simulations completed in minutes;
on-demand analysis concurrent with physical testing
Next Steps
The team has worked closely with the Department of Energy (DOE)
to perform initial demonstrations of the immense potential of
exascale-class computing for aerospace CFD, and is now bringing
this power to bear on challenge problems relevant to NASA missions
Human-scale Mars lander with retropropulsion
Aircraft in high-lift configuration
CFD Vision 2030 Roadmap Identifies HPC as Enabling For All Other Technologies
POC: Eric Nielsen, NASA Langley Research Center