Computer Science – Performance
Scientific paper
Dec 2010
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2010agufmsm31a1853w&link_type=abstract
American Geophysical Union, Fall Meeting 2010, abstract #SM31A-1853
Computer Science
Performance
[1900] Informatics, [1908] Informatics / Cyberinfrastructure, [1932] Informatics / High-Performance Computing, [1994] Informatics / Visualization And Portrayal
Scientific paper
In the Solar-Terrestrial Physics (STP) field, spatio-temporal resolution of computer simulations is getting higher and higher because of tremendous advancement of supercomputers. A more advanced technology is Grid Computing that integrates distributed computational resources to provide scalable computing resources. In the simulation research, it is effective that a researcher oneself designs his physical model, performs calculations with a supercomputer, and analyzes and visualizes for consideration by a familiar method. A supercomputer is far from an analysis and visualization environment. In general, a researcher analyzes and visualizes in the workstation (WS) managed at hand because the installation and the operation of software in the WS are easy. Therefore, it is necessary to copy the data from the supercomputer to WS manually. Time necessary for the data transfer through long delay network disturbs high-accuracy simulations actually. In terms of usefulness, integrating a supercomputer and an analysis and visualization environment seamlessly with a researcher's familiar method is important. NICT has been developing a cloud computing environment (NICT Space Weather Cloud). In the NICT Space Weather Cloud, disk servers are located near its supercomputer and WSs for data analysis and visualization. They are connected to JGN2plus that is high-speed network for research and development. Distributed virtual high-capacity storage is also constructed by Grid Datafarm (Gfarm v2). Huge-size data output from the supercomputer is transferred to the virtual storage through JGN2plus. A researcher can concentrate on the research by a familiar method without regard to distance between a supercomputer and an analysis and visualization environment. Now, total 16 disk servers are setup in NICT headquarters (at Koganei, Tokyo), JGN2plus NOC (at Otemachi, Tokyo), Okinawa Subtropical Environment Remote-Sensing Center, and Cybermedia Center, Osaka University. They are connected on JGN2plus, and they constitute 1PB (physical size) virtual storage by Gfarm v2. These disk servers are connected with supercomputers of NICT and Osaka University. A system that data output from the supercomputers are automatically transferred to the virtual storage had been built up. Transfer rate is about 50 GB/hrs by actual measurement. It is estimated that the performance is reasonable for a certain simulation and analysis for reconstruction of coronal magnetic field. This research is assumed an experiment of the system, and the verification of practicality is advanced at the same time. Herein we introduce an overview of the space weather cloud system so far we have developed. We also demonstrate several scientific results using the space weather cloud system. We also introduce several web applications of the cloud as a service of the space weather cloud, which is named as "e-SpaceWeather" (e-SW). The e-SW provides with a variety of space weather online services from many aspects.
Fukazawa Kenji
Inoue Susumu
Kato Harukazu
Kimura E.
Morikawa Yoshitada
No associations
LandOfFree
Demonstration of NICT Space Weather Cloud --Integration of Supercomputer into Analysis and Visualization Environment-- does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with Demonstration of NICT Space Weather Cloud --Integration of Supercomputer into Analysis and Visualization Environment--, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Demonstration of NICT Space Weather Cloud --Integration of Supercomputer into Analysis and Visualization Environment-- will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1470065