A Research On Cloud Storage Architecture Without Master Control

Internet is an information age. A lot of information blowout on the Internet and the applications of mass data processing will be the most important part in the future. How to store and manage the mass data effectively becomes one of the difficulties faced by the current network. Distributed cloud storage as a new distributed storage, with its flexible structure, high efficiency response and easy management features, becomes the enterprises and institutions' choice to solve the problem of explosive growth data. However, the study found that both traditional distributed file systems and distributed file system based on the Internet are far away from the cloud storage on the scalability, reliability, fault tolerance and other aspects. The reasons for these deficiencies are that all the systems use logic centralized architecture. And as the system size increasing, the disadvantages will become more apparent.Based on the above issues, this paper studied and summarized the current development of distributed storage system comprehensively. Taking a variety of cloud computing applications into consideration, this paper designed cloud storage architecture with uniting physical and logical. The architecture in this paper combined efficiency management of centralized storage and robustness of P2P storage, and changed the poor scalability of centralized management and the single point failure of master node. The architecture achieved the unity distributed of logic and physics and made the cloud storage systems with high scalability, reliability, user transparent and fault tolerance come true.Firstly, this paper designed cloud storage architecture without main control by uniting physical and logical. The architecture overcame the single point dependency of centralized architecture through the use of district self management SuperNode. Secondly, any application of centralized architecture can be provided by this architecture by SuperNode's collaboration, and the applications don't have any negative impact. Thereby the cloud storage architecture has high scalability. Finally, in order to achieve taking over the failure of a subgroup of SN automatically and quickly without any manual intervention, this paper designed a dynamic takeover mechanism. Through the improvement of DataNodes' heartbeat mechanism, the DataNodes under the fail SuperNode will be taken over dynamically and added to the new sub-group. So the cloud storage system has high fault tolerance. This paper's cloud storage architecture design is different from other traditional or existing storage architecture, and it's the real cloud storage. The cloud storage architecture in this paper laid a solid foundation for all sectors to storage and management the mass data effectively. This architecture can be implemented on common PC clusters, which greatly reduces the cost of implementing a distributed system.