The study on Adjustable-Threshold File Replication Strategy for Data Grids

• Main Authors: Hsuan-Fu Yu, 游瑄富
• Other Authors: Yi-Min Wang
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/31771908858404763937

碩士 === 靜宜大學 === 資訊管理學系研究所 === 98 === To shorten the file access latency and to reduce the network bandwidth consumption for data grid environments‚ many dynamic file replication strategies have been studied extensively recently. For example‚ Chang et al. proposed Latest Access Largest Weight (LALW) strategy for locality file access pattern on hierarchical data grid. Based on file access records, LALW first chooses the most popular file among replicated files‚ and computes the number of copies should be made for that popular file. Then LALW locates the suitable nodes for those copies. In this thesis‚ we propose a new dynamic file replication strategy‚ Adjustable File Replication (AFR)‚ to improve LALW. As the differences of accessed times among the leading popular files are insignificant. Instead of choosing only one popular file‚ AFR suggests that more popular files should be selected. Simulation results show that AFR improves the effective network usage and reduces the remote file access time on hierarchical data grids.