Details for NCL-TR-2010009
PropertyValue
NameNCL-TR-2010009
Description
An Adaptive Data Replication Algorithm based on Star-based Data Grids
Lee, Ming-Chang, Leu, Fang-Yie, & Chen, Ying-ping
Abstract: In recent Grid research and development, data replication has been used to duplicate frequently accessed data from its current location to appropriate sites so as to improve the whole system's data access performance and reduce bandwidth consumption for data delivery. Several data replication algorithms have been proposed. Some were designed based on unlimited storage. However, not all Data Grids are with unlimited storage space. Others were implemented on limited storage environments. However, none of the algorithms developed on limited storage environments has considered file popularity, defined as how often a file is accessed by users. In fact, file popularity and data access patterns of a system vary with time since users sometimes change their interests where data access patterns, defined as the distribution of access counts on files of a system, may influence on the data access performance of the system. In other words, the file replication model of the system might not be able to adapt to the change of users’ data access behaviors. Therefore, in this study, we proposed an adaptive data replication algorithm, called Popular File Replication First algorithm (PFRF for short), which is developed on a star-based Data Grid with limited storage space. With aggregated data access information of previously job execution and user behaviors, PFRF can predict future file popularity, and replicate potential popular files/replicas to appropriate cluster/sites to adapt the change. We employ several types of file access behaviors, including Uniform, Geometric, and Zipf-like distributions, to evaluate PFRF. The simulation results show that PFRF can effectively shorten average job response time, reduce bandwidth consumption for data delivery, and increase data availability as compared with the tested algorithms.
FilenameNCL-TR-2010009.pdf
Filesize2.47 MB
Filetypepdf (Mime Type: application/pdf)
Creatorypchen
Created On: 12/06/2010 18:14
ViewersEverybody
Maintained byEditor
Hits4265 Hits
Last updated on 12/09/2010 12:17
Homepage