基于混合策略的低成本云存储方案.pdfVIP

  • 8
  • 0
  • 约2.11万字
  • 约 7页
  • 2017-05-04 发布于北京
  • 举报
基于混合策略的低成本云存储方案.pdf

Journal of Computer Applications 计算机应用, 2014 , 34( 10) : 28∞ -2805 , 2811 ISSN 1∞1-9081 CODEN JYIIDU 2014-10-10 文章编号: 1∞1 -9081(2014)10 - 28∞ -06 doi: 1O.11772/j. issn.l∞1-9081.2014.10.28∞ 基于混合策略的低成本云存储方案 李松涛13· ,金欣3 (1.西北大学信息科学与技术学院,西安 71α)69; 2. 汉中职业技术学院机电工程系,陕西汉中 723∞0; 3. 清华大学深圳研究生院,广东深圳 518055) ( *通信作者电子邮箱 supertaozi@ 伊出1. com) 摘 要:为了保证云存储系统数据的高可用性、降低数据存储成本和带宽成本、缩短数据对象的访问时间,提出 一种称为缓存大小自适应确定(CAROM) 的新方案。 CAROM 结合传统的基于缓存策略的方法和纠错码方法来提高 云文件系统的弹性和效率。另外,为了在缓存大小及其效益问实现平衡,提出一种基于总体成本凸函数特性的自适 应方法来实现缓存大小的自适应选择。在基于现实世界文件系统数据的性能评估中,CAROM 方案的存储成本和带 宽成本分别比复制策略和纠错码策略下降ω%和 43;J毛,同时访问延时与复制策略相当。结果表明,CAROM 方案在支 持当前云文件系统语义一致性的同时,兼具带宽成本低、存储成本低和访问成本低等特性。 关键词:云计算;云存储;复和l策略;纠删码;访问延时;成本 中图分类号 :T凹91 文献标志码 :A Low-cost cloud storage scheme based on hybrid strategy HSongta01 ,2\JIN Xin3 ( 1. School 01 lr矿ÎJrmation Sc回nce and Techrwlogy, Noπhwest University, Xian Shaanxi 71∞69, Chinα; 2. Depart阳nt 01 Mechαnical Engineering, Hanzhong V.时ational and Technical College, Han.zhong Shaanxi 723∞0, China; 3. Cradua甜 Sc仇。1 01 Shenzhen, Tsinghua Uni时rsity, Shenzhen Cuangdong 518055 , China) Abstract: To ensure the high data availahility in cloud storage system, reduce the cost of data storage and bandwidth, shorten the time for accessing dala objects, a new scheme named Cache A Replica On Modification (CAROM) was proposed. It combined traditional methods based on replication strategies and erasure coding strategies to improve the flexibility and efficiency of cloud file syslem. In addition, in order to achieve a balance between the cache size and efficiency, an adaptive method based on the characteristics of 由e overall cost of a convex function was proposed to select adaptive cache size. A large- scale evaluation using real-world file system was conducted, and CAROM outperformed replication based schemes in slorage cost by up to 60% and erasure coding schemes in bandwidth cost by up to 43 ;J岛, while maintaining low access latencies close to those in replication based schemes. The r

文档评论(0)

1亿VIP精品文档

相关文档