[P-EM35_2PM2] Plasma Astrophysics: simulation technique, method of data analysis/visualization

[PEM35-06] A Study of Fluid Element Tracing in Global MHD Simulations via Parallel Data Processing on the NICT Science Cloud

*Ken T. MURATA¹, Keiichiro FUKAZAWA², Zhihong GUO³, Yusuke EBIHARA⁴, Satoko SAITA⁵, Shigeru FUJITA⁶, Takashi TANAKA⁷, Yasubumi KUBOTA¹, Hidenobu WATANABE¹, Osamu TATEBE⁸, Masahiro TANAKA⁸, Eizen KIMURA⁹ (1.National Institute of Information and Communications Technology, 2.Kyushu University, 3.SURIGIKEN Co.,Ltd., 4.Kyoto University, 5.Research Organization of Information and Systems, 6.Meteorological College, Japan Meteorological Agency, 7.SERC, Kyushu University, 8.University of Tsukuba, 9.Ehime University)

The NICT Science Cloud is a cloud system designed for scientific researches, and expected as a new infrastructure for big data sciences. Not only parallelization of CPU as in super-computers, but I/O and network throughput parallelization are crucial for the big data science. A high-performance visualization system is constructed on the NICT Science Cloud using Gfarm/Pwrake middleware. We examined performance of this parallel visualization environment for a set of computer simulation with 1000 files (2.3TB in all). After setting higher priority to access to local file on local disk, we finally achieved 124 times higher visualization using 192 core cpu.