| コモディティグリッド技術によるテラスケール大規模数理最適化
|
| | | Abstract: We have constructed the TeraFlops test-bed, 886 processors and 3 or more TeraFlops by the participating research organization (Tokyo Institute of Technology, Tokyo Denki University, and University of Tokushima) total, using the commodity grid technology by the PC technology and the high-speed network, and researched the fault tolerant and scalability of Ninf-G which is a GridRPC system on our test-bed. On our test-bed, we have examined the scalability and performance of our various software of mathematical optimization. Our various software of mathematical optimization is controlled by Ninf-G between each organization and by Ninf-1 in each organization. In an experiment, it was able to succeed in the parallel computing of a maximum of 770 processors among 886 processors, and a scalability and stability were able to be shown. | | | |