The Research On Data Replica Placement In Geo-distributed Cloud Storage System

With the development of internet technology,cloud computing has been more and more widely used,and the amount of data has generated shows an explosive growth.How to store mass data effectively has always been a research hotspot.In recent years,especially with the rapid development of big data and cloud computing,a single data center cannot meet the current demand.It has one solution that deploys the data center in a Geo-distributed environment;deploy multiple data centers in different geographical locations at the same time.It's very important that how to ensure the performance of Geo-distributed system.The reasonable placement of data replica in the system is a method that can effectively improve system performance.The main work of this dissertation includes the following contents:(1)Based on the traditional replica placement method and the characteristics of Geo-distributed environment,a new method of replica placement has been proposed,which has not only considers the current storage load of the data node,but the factors of performance in the data node such as the network bandwidth,the disk load and the data node distance.(2)A geo-distributed data replica placement method based on hot of access has been proposed in this dissertation.When computing the visiting rate of data access,it use dynamic time window which can adjust the size of time window according to the change of visiting rate.Considering the difference of data nodes in the Geo-distributed environment,the replica of high-visiting rate are placed into the data nodes of high performance according to the historical information,which can ensure the performance of the systemand improve user access efficiency.Finally,It has designed a prototype system for Geo-distributed cloud storage that based on the Hadoop,and replica replacement method have been implement in this system.The experiment results show that the proposed data replica replacement method is feasible.