宇宙機開発における大規模CFD解析結果の遠隔並列可視化(宇宙科学情報解析論文誌第一号)

概要

論文の詳細を見る
PC cluster is installed to evaluate the parallel and remote visualization of large-scale computational results produced by the supercomputer in the remote site. It is revealed that more than 60 % of the time for visualization is required for reading the binary data of the computational result. Partitioned file format such as VTK multi-block is desirable for reading large-scale data in parallel. High scalability more than 80 % is obtained by using the parallel rendering and image compositing. Methodology to extract the region of interest (ROI) based on the VTK multi-block format file is developed, and reduction of reading data size is realized for large-scale computational result without losing the parallel I/O performance. The knowledge obtained here will be utilized for the post-processing system of the large-scale computations done by remotely connected supercomputers.