The International Conference for High Performance Computing, Networking, Storage, and Analysis
The International Conference for High Performance Computing, Networking, Storage, and Analysis
Student: Daoce Wang (Washington State University)
Supervisor: Dingwen Tao (Washington State University)
Abstract: Large-scale cosmology simulations generate large amounts of data for post-analysis, resulting in I/O and storage bottlenecks. This work investigates an effective in-situ error-bounded lossy compression for Nyx, an adaptive mesh refinement (AMR) based cosmology application. Our contribution is threefold: (1) We explore the best-fit in-situ error-bounded lossy compressor (including SZ and TTHRESH) for Nyx considering both compression ratio and post-analysis quality. (2) We propose an approach to adaptively optimize the compressor for different AMR levels based on our developed metric and data characteristics. (3) Our evaluation shows that our approach can improve the compression ratio by 1.7X over the baseline (i.e., from 66 to 116) with the same post-analysis quality.
ACM-SRC Semi-Finalist: no
Poster: PDF
Poster Summary: PDF
Back to Poster Archive Listing