Simulations performed on Summit, the nation’s fastest supercomputer, are now publicly available to the scientific community

Scientists at NASA aim to one day land giant payloads on Mars, but safely descending passengers, flight crew, and other equipment in an atmosphere much thinner than Earth’s poses a significant challenge. Scientists studying potential Mars landing technologies need large-scale computational resources to model such scenarios, and to accurately model different technologies, they need new tools that can only be developed from realistic data sets.

A team at NASA running code on the nation’s fastest supercomputer, Summit at the US Department of Energy’s (DOE’s) Oak Ridge National Laboratory (ORNL), has made six data sets publicly available to aid this effort.

The team published nearly 400 terabytes of data characterizing six different flight scenarios for a deceleration concept known as retropropulsion, a descent technology that scientists aim to use for future human missions to Mars. Using retropropulsion for a Mars landing would involve firing a spacecraft’s engines to oppose the motion of the vehicle, which would cause it to decelerate as it entered the Martian atmosphere. The data sets are available to the community in the FUN3D Retropropulsion Data Portal.

“Since engineers have limited understanding of the complex plume structures involved, it is critical to be able to visualize the data in many different ways,” said Patrick Moran, a scientific visualization expert at NASA Ames who helped analyze the simulations. “Visualizing data sets of this size and complexity can be incredibly challenging. Making the data publicly available provides realistic test cases for prototyping of new tools and algorithms.”

The ability to visualize retropropulsion can lead engineers to new insights from nuanced details captured much more easily in high-resolution images. New insights could inform future studies, shape future spacecraft design, and provide stakeholders with information that could guide funding decisions.

Supersonic retropropulsion: using rocket motor thrust in the direction of travel to slow a vehicle entering the atmosphere of Mars. In this simulation, the vehicle is traveling at Mach 2.4, more than twice the speed of sound. The rocket motor plumes are rendered using a translucent isosurface of total temperature. Color is used to map the coefficient of pressure on the vehicle surface, where red regions and blue regions correspond to areas where the pressures are higher and lower than ambient, respectively. Video Credit: NASA

The cases in the data portal characterize a spacecraft flying at three different Mach numbers, or how fast the vehicle is traveling relative to the speed of sound. Each Mach number scenario includes a high- and low-resolution case for a total of six cases.

The team spent several years extending the computational fluid dynamics (CFD) code FUN3D to GPU architectures to be able to perform the simulations necessary to generate the data sets. CFD codes are used to simulate the motion of fluids (gases and liquids), such as those encountered during a Mars descent. Mike Matheson, visualization specialist in the Advanced Data and Workflow Group at the Oak Ridge Leadership Computing Facility (OLCF), served as a liaison on the project. The OLCF is a DOE Office of Science User Facility located at ORNL.

The amount of project data was so large that the team considered trucking it across the United States—from ORNL in Oak Ridge, Tennessee, to NASA’s archival storage at the NASA Advanced Supercomputing (NAS) facility at the Ames Research Center in Silicon Valley—in the form of physical tape drives. However, with the help of network experts at each facility, the team was able to transfer several petabytes of data across the internet at a sustained rate of 40 terabytes per day. As NASA’s Paul Kolano explained, “The NAS-developed Shift automated transfer tool was specifically designed to support these types of large cluster-to-cluster remote transfers. After tweaking transfer parameters to maximize performance, the multiple levels of parallelism within Shift allowed the transfers to not only be carried out in many-to-many fashion at high speed but, just as importantly, did so while ensuring the full end-to-end integrity of the data.”

“We really wanted to publicly share these data sets so that we could give an equal opportunity to everyone to work with them,” Moran said. “We want to encourage as much participation as we can so that we both gain further insight into the current study and encourage the development of techniques that will help us get the most from future studies.”

The research team include Ashley Korzun, Eric Nielsen, Aaron Walden, Jan-Renee Carlson, and Bill Jones of NASA Langley; Matheson of ORNL; Chris Henze, Kolano, Moran, and Tim Sandstrom of NASA Ames; Justin Luitjens of NVIDIA; and Mohammad Zubair of Old Dominion University.

UT-Battelle LLC manages Oak Ridge National Laboratory for DOE’s Office of Science, the single largest supporter of basic research in the physical sciences in the United States. DOE’s Office of Science is working to address some of the most pressing challenges of our time. For more information, visit