Research Computing helps client run large simulations
Customer Case Study
Tim Weigand is a Ph.D. candidate and research assistant at the Department of Environmental Sciences and Engineering, within the Gillings School of Public Health at UNC-Chapel Hill. Weigand is a client of ITS Research Computing.
Weigand has been working with his adviser, Casey Miller, on studying how freshwater, saltwater and sand interact as an application of a porous media problem. A porous medium is a material with pores that fluid, either liquids or gasses, may pass through. In this case, the fluid passing through the medium is water of different salinities. The goal of this research is to gain a deeper understanding of the mechanics and functions of coastal environments, particularly in the context of rising sea levels.
Weigand had been creating digital simulations of units of a thousand grains of sand and fluid to study how the materials would mix. The purpose of creating these simulations is to have very exact data and behavior that can be used to derive equations that are applicable to much larger, more realistic environments. The research group focuses on applications of mathematics and physics to environmental science questions.
Originally, Weigand had run his simulations on the supercomputers at Oak Ridge National Laboratory in Tennessee. When the supercomputer he was using was no longer available, Weigand had to find an alternative. He reached out to Research Computing to inquire if the Dogwood computing cluster would be a viable option for his computations. Every job that Weigand runs outputs more than 2 terabytes of data.
Research Computing made temporary allowance to the Dogwood computing cluster, which is a resource shared by many researchers across the UNC-Chapel Hill campus, to accommodate Weigand’s needs. At Oak Ridge National Laboratory, which is a U.S. Department of Energy research facility, Weigand was accustomed to a much larger computing capacity.
“Tim’s workload is much larger than Research Computing usually provides for,” said Liam Greenwood, Research Computing’s IT Manager for Advanced Computing Infrastructure. “When Tim came to us with his problem, we were able to make a special temporary reservation of 48 dedicated servers with 2112 CPUs and nearly 25TB of memory.”
Research Computing Administrator Jenny Williams has provided invaluable assistance to Weigand by tuning his workflow. This has enabled Weigand’s jobs to effectively utilize the resources.
Using a combination of OpenFOAM, a computational fluid dynamics software, and his own code, Weigand is able to run the larger simulations through the processors at the Dogwood cluster. Individual jobs take about three days to run. Weigand’s queue has been running continuously throughout the month of July.
“Tim has had the use of these resources since the beginning of June, and he is expected to complete his computations in August,” Greenwood said. “His need exceeded the typical use of this shared resource, and Research Computing is pleased we were able to manage the resources to enable completion of this specific important workflow.”
Weigand has been successfully using the Dogwood computing cluster to process his research data.
“Doing this research computationally rather than in a lab is important because we can control individual variables to isolate unique phenomena,” said Weigand, explaining the benefits of the digital simulations. “There have been a lot of attempts to generate a macro scale model as existing models could be stronger.”
Processing large data sets is a key part of many kinds of research. Because UNC-Chapel Hill is a research university, it is important for Research Computing to provide this kind of computing resource to faculty and students. Doing so supports the important research that Carolina’s faculty, staff and students pursue, and enables them to expand upon what is possible.