Performance of scientific simulations that are executed on large-scale computers is highly dependent on the volume of data transferred and the available bandwidth to transfer the data during the simulation. Higher performance decreases the time necessary for obtaining results from a scientific simulation and enables faster scientific advancement. This project seeks to understand the trade-offs of performance and accuracy when using lossy compression to reduce the volume of data transferred during scientific simulations. Lossy data compression significantly reduces the volume of scientific data by trading a loss in accuracy of the decompressed data for larger reductions in the compressed data size. The goal of this project is to investigate how lossy compression impacts the fidelity of the results of scientific simulations, which is critical for understanding how lossy compression can be effectively applied in practice. This project fosters the use and importance of large-scale computing resources for undergraduate STEM students and includes research experiences for undergraduates who are actively engaged in research tasks.
This project considers important trade-offs in using lossy compression for scientific simulations such as: energy efficiency, sensitivity of compression and decompression time and compression ratio to the selection of compression parameters, and impact on statistical measures. This project analyzes the effects of lossy compression on real-world scientific data from a diverse set of domains such as fluid dynamics, astrophysics, and genomics. Understanding these trade-offs guides the integration of lossy data compression into computational kernels and applications. This project is jointly funded by CCF/SHF core program and the Established Program to Stimulate Competitive Research (EPSCoR).
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
