The Research And Design Of Distributed Network File Storage System

With the fast development of information network and Internet, our life gets more and more intelligent. In face of a large amount of Internet information everyday, there is an explosion of personal data, so how to meet the demand of massive data storage is an urgent agenda. However, traditional storage methods cannot meet the demand because of its problems, such as insufficient open, high cost of construction, and poor scalability. As a result, the development of cloud computing brings a great opportunity for cloud storage, that is, there is no need for the intelligent terminals to have large storage space, instead, the personnal data is migrated to be stored in cloud, and the data can be visited by users anytime and anywhere through the Internet. In a matter of speaking, the development of cloud storage is overthrowning the existing framework of network storage.This thesis presents a solution for the storage of network files according to the need of intelligent termimal users. A distributed storage system for nentwork files based on MooseFS is designed and implemented, and a model based on dynamic replica mechanism and capacity-aware mechanism is constructed, and all of these will provide safe and transparent network storage for the users.Firstly, network disk is designed according to the demand of storing in cloud. It is supported by the LAMP framework, and assigns personal storage space for each user. Users can manage their own files, such as uploading and downloading. So the network disk realizes the migration of storage from terminals to cloud, and it provides a visual entrance of visiting the storage space in cloud.Secondly, to satisfy the requirement of massive storage and to promote the fatety of storage, the network disk is designed to store files distributedly. One scheme is to make use of the distributed file system with open source, and MooseFS is chosen in this thesis. That is, construct the server cluster of MooseFS, and mount the storage space of network disk on the MooseFS. By this, it can realize the distributed and transparent storage of files. Another scheme is to design and construct specific distributed storage model for massive small network files. In the proposed model, JNotify is used to implement realtime monitoring. And dynamic replica mechanism is introduced to make the replica of files positive correlation with the use frequency, and this can make a trade off between storage reliability and utilization rate of the resources; and capacity-aware mechanism is introduced to make the nodes availability positive correlation with the actual load rate, and this can promote the utilization rate of node capacity. Meantime, back management programme is developed based on MVC to make the manipulation of the model visual, and it makes the model management more convenient.To solve the contradiction between the limited storage space of terminals and the requirement of massive data storage, distributed network storage system with MooseFS is realized, and build up specific distributed storage model for massive small network files, the results show that the data distribution strategy of the model has better feature in capacity awareness and resources utilization. The solution proposed in this thesis means a lot to the improvement of the quality of life and the satisfaction of the informatization development.