Research On Replica Management Strategy In Cloud Storage Of Multi-Data Center

With the arrival of "big data" time, in which there are massive amounts of data and things are all digitized, cloud storage will play an increasingly important role in the storage and management of these massive data. In the cloud storage system, data replication technology is often used to improve the throughput, reliability, security, availability and scalability of the system. But the research of replication technology in cloud storage environments is more confined to a single data center at present. From the perspective of disaster recovery and load balance, the replication management strategies on multiple data centers need to be studied []].Replication management of cloud storage environments which is on multiple data centers need to face new problems, such as:how to choose a the data center to place the replicas, how to reduce the data transmission across the data centers, how to maintain the consistency of replicas across data centers, and how to both improve data reliability and system load balance, taking into account the good user experience Although existing cloud companies have established some multi-data center cloud storage systems, few details of their technologies have been revealed. In this thesis, focusing on data management strategies under multi-data center cloud storage environments, I have done the following research and exploration work:1. Took a comprehensive understanding of cloud storage system, multiple data centers and replication management technology, focused on the replication management strategy in the typical cloud storage system HDFS. Then I analyzed the problems of multi-data center cloud storage system replica management facing;2. For no practical and open source multi-data center cloud storage systems are available, based on HDFS system, I proposed a two cascading master-slave storage system architecture for multi-data center cloud storage system2CMSSS (2Cascading Master-Slave Storage System), which simplified multi-data center cloud storage system research work, and laid the foundation for studying the multi-replica management strategy of multi-data center storage system and the subsequent expansion and improvement;3. Based on AHP-backward cloud algorithm, I proposed the data center selection strategy.Since the dynamic network environment and the differences between the data centers, this thesis presented data center selection strategy based on AHP-backward cloud algorithm.This data center selection strategy paved the way for proposing the data placement strategy in multi-data center cloud storage system environments;4. Combined with user access of cloud storage system and the data center selection strategy proposed in the front of this thesis, I proposed the data placement strategy in a multi-data center cloud storage system based on user-driven and data center selection strategy(Data Center Select-User Drive,DCS-UD). Simulation results show that, the data placement strategy in multi-data center cloud storage system,which is based on data center selection strategy with user-driven, have a great advantage in system load balance, replica access rate and data reliability;5. Maintain the consistency of replicas is a critical issue for multi-data center cloud storage system. Based on2CMSSS system model and locking mechanisms, I summed up the design goals and ideas of replica consistency in multi-data center cloud storage system,and proposed the replica consistency strategy of multi-data center cloud storage system based on the2CMSSS and two phase locking mechanism. The simulation shows that the strategy can ensure the replica consistency and lower user access latency...