The Methods And Optimizations For Mass Data P2P Distributed Steady Storage

The mass data P2P distributed steady storage architecture is based on the existing P2P technology, which is composed with large amount of cooperative P2P storage nodes and according to the system service principle of resource sharing for store data. In order to achieve the organic composition of the distributed high-reliable P2P storage architecture, the P2P distributed storage system has to face with a highly dynamic environment of network storage. Each P2P storage node has its own user behavior features and nonlinear online/offline changes independently. To research such a high degree of complexity of large-scale distributed storage system for steady storage is a challenge, and will cause many storage management problems that the traditional distributed storage architecture is difficult to solve effectively.The thesis, around with the mass data P2P distributed steady storage architecture, focuses on the storage structure programming and the storage organization management. The goal of the thesis is to set up a P2P distributed storage system with nice storage capacity under the existing network conditions. The mass data P2P distributed steady storage problem is a comprehensive issue, it's facing the demand of mass data steady storage and the condition of complexity storage environment at the same time, which constructs a both interrelated and mutual constrained system research mode. By drawing on the research results of the traditional distributed storage systems and combining the new features of the P2P technology, the thesis is committed to the reasonable storage architecture and its management optimizations. The methods make up the contradictions between loose P2P storage organization and steady data storage request, which is validated by simulations and cases. The main aspects of the thesis include as follows:(1) Researching on the overall concept of mass data P2P distributed steady storage, the thesis proposes that the steady storage has three basic characteristics of capacity according to the system running features: the durability of data existent, the stablility of data access and the survivability of data sudden attack. Then, based on the capacity division, the thesis establishes different reliability models to reflect the steady storage concept and selects approperate reliability indexes to quantify the basic characteristics, which act as reference models for all over the thesis.(2) Researching on the structure programming of mass data P2P distributed steady storage problem, the thesis adopts reasonable structure for P2P data redundancy storage in order to improve the overall fault-tolerant level of the storage system. In the thesis, the structure programming includes two specific elements: Firstly, in the way of the P2P redundancy storage, the thesis compares the mode of complete duplication redundancy and erasure code redundancy roundly, and chooses the MDS erasure code, with the best fault-tolerant capacity, as the redundancy storage method for steady storage. The thesis creates a performance function for storage– management cost to set the redundancy parameters of the MDS erasure code redundancy storage, and construts a systemetic MDS erasure code with n value extension to improve the storage capacity of the P2P rendundancy storage. Secondly, in the way of the P2P storage structure, the thesis analyses the performance degradation of the P2P erasure code rendundancy storage structure, and proposes an appropriate P2P segmentation erasure code redundancy storage structure called SegEra. The thesis creates an effective progarmming model for the SegEra's segmentation size, and achieves the best SegEra structural parameters for the P2P distributed storage system.(3) Researching on the organization management of mass data P2P distributed steady storage problem, the thesis adopts self-adaptive management for P2P data redundancy storage in order to suit the complex collaborative storage environment. In the thesis, the management optimization includes three specific elements: Firstly, in the way of the P2P failure maintance mechanism, the thesis considers the behavior features of the P2P temporay and permament store data failure, and proposes a tradeoff P2P failure data maintance method for determine accuracy and recovery speed. The thesis uses SWT (Stationry Wavelet Transform) and BPNN (BP Neutral Networks) to forcast the expectation of the P2P storage node's offline time according to its history records, and calculates the optimal latency time before repairing the P2P failure data by the recovery efficiency model. Secondly, in the way of the P2P hierarchical organization mechanism, the thesis emphasizes the difference in strong/weak P2P storage node architectures, and uses the TAM (Technology Acceptance Model) to describe the change law of the P2P storage node's ability with CA (Cellular Automata) evolution analyses. According to the node's ability change law, the thesis divides the P2P nodes into three levels: united storage level, cooperative storage level and shared storage level. With different storage management methods, the hierarchical P2P storage organization shows an improvement effect in the P2P storage experiments. Thirdly, in the way of the P2P survivability distribution mechanism, the thesis discusses the deficiencies of the general store data distribution methods, and introduces the network survivability conditions for the factor of the P2P store data distribution considerations. Based on the classical network survivability theory, the thesis creates a P2P store data distribution model to associate the storage survivability and data redundancy distribution, and designs the store data survivability measures for data storage topology to optimize the P2P store data survivable distribution method.