Presentation 2010-06-28
Optimizing Stream-based Anomaly Detection Algorithm SST with GPU
Kosuke MORITA, Toyotaro SUZUMURA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper we describe the implementation and evaluation of an anomaly detection algorithm called SST(Singular Spectrum Transformation)on top of a data stream management system, System S developed by IBM Research. Moreover we optimized the program in such a way that the most dominant computation part, SVD(Singular Value Decomposition)is offloaded to GPGPU. In the case of the short rage of time-series data, the performance of CPU-based approach outperforms the one of GPU-based approach, but when the window size becomes large, GPU achieves less than 5 seconds while CPU-based one takes nearly 1 minutes.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Data Stream Processing / DSMS / DSPS / System S / SPADE / GPGPU / Anomaly Detection / SST
Paper # DE2010-4
Date of Issue

Conference Information
Committee DE
Conference Date 2010/6/21(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Data Engineering (DE)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Optimizing Stream-based Anomaly Detection Algorithm SST with GPU
Sub Title (in English)
Keyword(1) Data Stream Processing
Keyword(2) DSMS
Keyword(3) DSPS
Keyword(4) System S
Keyword(5) SPADE
Keyword(6) GPGPU
Keyword(7) Anomaly Detection
Keyword(8) SST
1st Author's Name Kosuke MORITA
1st Author's Affiliation Tokyo Institute of Technology()
2nd Author's Name Toyotaro SUZUMURA
2nd Author's Affiliation Tokyo Institute of Technology:IBM Research-Tokyo
Date 2010-06-28
Paper # DE2010-4
Volume (vol) vol.110
Number (no) 107
Page pp.pp.-
#Pages 6
Date of Issue