A Customer Experience-Based Cost Minimization Strategy of Storing Data in Cloud Computing
Ning WANG,Yang YANG,Kun MENG,Yu CHEN,Lei WANG,Qing JI
DOI: https://doi.org/10.3969/j.issn.0372-2112.2014.01.004
2014-01-01
Tien Tzu Hsueh Pao/Acta Electronica Sinica
Abstract:Splitting a whole large file into some small blocks and distributedly deploying these block replications on severs can improve their application efficiency and meet users' demands. How to split a file, how many replications of a block should be created, and how to deploy these replications become the most critical issures in this field. In this paper, considering client's demands for each block, we introduce block-popularity and present a block deployment strategy according to it, called minimum cost of storing blocks (MCSB), which can satisfy users' requirement and minimum storage cost. In our strategy, through introducing cost adjustment factors of blocks, not only that lower hot blocks are assured to deploy earlier, but also the practical total storage cost are not exceed than the original case. To maintain the performance of service system, we study resource management methods in-depth among severs with different workloads and propose an adaptive MCSB strategy (AMCSB). We implement the MCSB (and AMCSB) in Hadoop and compare their performance with other related strategies. The experiment results show that MCSB fulfills minimum service cost and shortens average response time, and AMCSB realizes load balance among target servers.