Search Problem Solving Based on Credit Evaluation for Volunteer Computing

Takuya IGUSA  Tomofumi MATSUZAWA  Masayuki TAKEDA 

Publication
B - Abstracts of IEICE TRANSACTIONS on Communications (Japanese Edition)  Vol.J95-B  No.7  pp.827-834
Publication Date: 2012/07/01
Online ISSN: 1881-0209
Print ISSN: 1344-4697
Type of Manuscript: Special Section PAPER (Special Section on Frontiers by Young Researchers)
Category: 
Keyword: 
volunteer computing,  public resource computing,  parallel computing,  clustering, 

Full Text(in Japanese): PDF(1.7MB)

Summary: 
We propose a method of solving a search problem efficiently for volunteer computing (VC) environment. In a conventional master worker model like BOINC, when a node interfered with VC projects or went wrong, there was a problem that the reliability and processing efficiency of a calculation result fell. In our method, more efficient VC systems can be built by clustering nodes based on the credit and distributing the optimal task and redundant calculation according to the past percentage of correct answers. Simulation of solving search problems indicates that our method improves reliability and efficeincy.