| Decentralized system is an open system composed of highly autonomous nodes which are connected freely.Centralized system has many problems,such as high cost of central server,over-dependence on backbone network and single point of failure.Compared with the centralized system,the decentralized system has the advantages of low cost,distributed data storage and permanent storage.But the benefits of decentralized technology come with the following challenges:First,the physical distance between nodes in the decentralized system does not match the logical distance,resulting in the waste of bandwidth in data transmission;Second,the decentralized system is based on content addressing,so the data query is time-consuming and inefficient;Thirdly,There are hot data problem in decentralized system,and the traditional load balancing method is not suitable for the decentralized system.Aiming at the above challenges,this thesis proposes a decentralized data organization mechanism based on organizational hierarchy and a corresponding data query and retrieval algorithm.At the same time,the solution to the hot data in the central system is proposed.The main work of this thesis is as follows:(1)Propose decentralized data organization mechanism and data auxiliary backup mechanism based on organizational hierarchy.Based on the hierarchical and decentralized data organization mechanism,the data query efficiency is improved through the physical distance priority query and the subordinate organization priority query.The data auxiliary backup mechanism ensures the backup of node data in the unreliable network environment.(2)A decentralized data query algorithm based on binary tree dynamic Bloom filter is proposed.In order to solve the problem of time-consuming data query caused by lack of centralized index in decentralized system,a decentralized data query algorithm based on dynamic Bloom filter is proposed to optimize the data query path by using the neighbor information saved by experienced nodes in the process of data query.In order to solve the problem of high false positive probability of Bloom filter,an improved dynamic Bloom filter data query algorithm BTDBF based on binary tree is proposed.(3)In view of the hot spots of data index and data service existing in decentralized data organization,a decentralized data load balancing mechanism based on the heat model is proposed.In order to solve the problem that the index of decentralized system is clustered in a specific node,which leads to the hot spot of data index,a data resource publishing and query mechanism based on index hot spot detection is proposed.In order to solve the problem that different nodes bear different data requests due to the heterogeneity of nodes,a data heat monitoring method based on node load is proposed,and a data service hot spot processing mechanism is implemented to cooperate with other nodes.In order to verify the effectiveness of the above work,this thesis conducts functional and performance experiments on decentralized data organization mechanism based on organizational hierarchy,data query algorithm based on BTDBF and two kinds of hot spot solutions respectively.The experimental results show that the decentralized data organization mechanism and data query algorithm proposed in this thesis has a good improvement in query performance compared with the IPFS file system,and the two kinds of hot spot solutions can effectively solve the hot spot problems. |
PDF Full Text Request