Research Of Repository Replication And Selection Strategy And Model Of Personalized Cloud Synchronization System

With the explosive growth of network information data, cloud storage on the basis of the rapid development of cloud computing, the purpose is to businesses and individuals can be freed from the heavy and complex data store. Due to the complexity of network data and the huge of enterprise data storage,it requires cloud storage system should have good reliability, availability, expansibility to make users have a good experience when they use it.Save multiple copies of the data is an important way to make cloud storage meet the above requirements. Effective replication strategy is an important research topic of cloud storage system,and the number of replication, the replication layout, the cost control between replications synchronization should be studied in deeply.This work studies selection strategy and model research of repository replication based on personalized cloud synchronization system,focused on solving the replica creation,deletion,selection, redundancy, layout and so on. The main work is summarized as follows:Firstly,the work analyzes the status of cloud storage and summarizes following problems with the cloud storage software:First, the function of mobile terminals app too simple to meet the modern demand for anytime, anywhere access to information; Second, many the terminal cloud storage softwares do not have the record management function, it is difficult for users to get historical versions of files. Based on the above two issues, personalized cloud synchronization system which controls the file versions using on the mobile terminals is designed. The mechanism is divided into three, GIT protocol storage layer, transaction layer cloud-based platform, and mobile client layer.The most important part of this theory is the strategies and models of personalized cloud synchronization system,which are related with creating and choosing a copy of the repository. Getting the idea from the traditional repository replication in Hadoop Distributed File System, the paper presents a dynamic duplication technology with GIT repository heat and create a policy model, than give its calculation method of GIT repository heat, maximum number, minimum number and the process of selecting the replication of Buffernode based on Topsis algorithm in detail.Next aiming at copies of the optimal problem in the process of cloud synchronization, the paper put forward the concept of the environment similarity and the Buffernode credibility. Uses the analytic hierarchy process (AHP) to determine the weight each influence factors in the Buffernode credibility, and gives the steps of the algorithm.On the basis of the above theory, the author begins experimental simulation in two ways for comparison.One is introduced in third and fourth paragraph,and the other is in traditional HDFS.Compare the experimental results,and obtain the conclusion.The system in this paper can handle more per unit of time access service. The system based on environment similarity and Buffernode credibility are more stable in customer service. Finally on the results of the analysis, give the corresponding improvement scheme and future development prospects in the subject...