Home >

news Help

Publication Information


Title
Japanese:データストリーム処理におけるGPUタスク並列を用いたスケーラブルな異常検知 
English:Scalable Anomaly Detection on Data Stream Processing with GPU Task 
Author
Japanese: 上野 晃司, 鈴村 豊大郎.  
English: Koji Ueno, Toyotaro Suzumura.  
Language Japanese 
Journal/Book name
Japanese:先進的計算基盤システムシンポジウム論文集 
English:Proceedings of Symposium on Advanced Computing Systems and Infrastructures 
Volume, Number, Page         pp. 193-200
Published date May 9, 2012 
Publisher
Japanese: 
English: 
Conference name
Japanese:SACSIS2012 - 先進的計算基盤システムシンポジウム 
English:SACSIS2012 - Symposium on Advanced Computing Systems and Infrastructures 
Conference site
Japanese:神戸 
English:Kobe 
Abstract Stream computing has emerged as a new processing paradigm that processes incoming data streams in a real-time fashion. On the other hand, many recent efforts have shown the suitability of GPGPU to high performance computing. By bringing two new trends, this paper proposes new innovative method called GPU task parallelism to optimize stream computing with GPGPUs. In this paper we implement the proposed approach over SVD (Singular Value Decomposition) and IKA-SST, a powerful algorithm of change point detection. The experimental results show that the proposed implementation of SVD provides performance gain by around 4 times order against quad-core and the proposed implantation of IKA-SST provides around 20 times order against single-core. This result validates the scalability of our proposed approach.

©2007 Tokyo Institute of Technology All rights reserved.