Presentation 2010/7/28
Utilizing GPU nodes for Low-power Data Intensive Computing
Yoshiyuki OHNO, Dai KOBAYASHI, Masaki KAN,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Information-explosion in the real world demands leveraging huge data sets by high speed and adequate energy cost. The objective of our research is realizing Data Intensive Computing System, which manages both high-performance and energy-saving, by utilizing General purpose GPU. We propose strategies of allocating data and task, contolling states of data-computing nodes in distributed system. The evaluation results on a GPU/CPU mixed cluster show the elimination of data I/O competition to control data allocations before task-executions. The results also show that this proposal strategy can increase 1.36x performance and save 20.7% of energy consumption compared with execution on only CPU nodes.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Data-intensive Computing / GPGPU / Energy-saving
Paper # CPSY2010-8
Date of Issue

Conference Information
Committee CPSY
Conference Date 2010/7/28(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 Computer Systems (CPSY)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Utilizing GPU nodes for Low-power Data Intensive Computing
Sub Title (in English)
Keyword(1) Data-intensive Computing
Keyword(2) GPGPU
Keyword(3) Energy-saving
1st Author's Name Yoshiyuki OHNO
1st Author's Affiliation System Platforms Research Laboratories, NEC Corporation()
2nd Author's Name Dai KOBAYASHI
2nd Author's Affiliation System Platforms Research Laboratories, NEC Corporation
3rd Author's Name Masaki KAN
3rd Author's Affiliation System Platforms Research Laboratories, NEC Corporation
Date 2010/7/28
Paper # CPSY2010-8
Volume (vol) vol.110
Number (no) 167
Page pp.pp.-
#Pages 6
Date of Issue