标题：DATA REPLICA'S DISTRIBUTION OPTIMIZATION BASED ON AVAILABILITY PREDICTION USING FUZZY THEORY IN THE CLOUD
作者：Wu, Xiuguo; Jiang, Xuejun
作者机构：[Wu, Xiuguo; Jiang, Xuejun] Shandong Univ Finance & Econ, Sch Management Sci & Engn, Jinan 250014, Shandong, Peoples R China.
会议名称：4th International Conference on Fuzzy Systems and Data Mining (FSDM)
会议日期：NOV 16-19, 2018
来源：JOURNAL OF NONLINEAR AND CONVEX ANALYSIS
关键词：Cloud; replicas distribution; availability; fuzzy; Qos
摘要：Cloud storage system has more advantages than traditional data processing mode, such as high scalability, high fault tolerance and huge storage space, where data replica is a common used technology aiming to enhance data access reliability and availability. However, traditional replicas distribution strategies mostly focused on current environment, usually leading to lower data availability in the near future. To solve this problem, a data replicas availability prediction model based on fuzzy theory is first presented using information of CPU, bandwidth, memory and security as input, while data replicas availability as output. After that, a two-phase data replicas distribution optimization algorithm based on the fuzzy availability prediction is proposed, including data replicas number and their placements. Simulations and experiments are performed to evaluate the validity and reliability of proposed approach, and results show that replicas space of data replica distribution optimization strategy accounts for nearly 70 percents compared with other two strategies, and less wait latency can be achieved through it, indicating the better effectiveness and practicality.