Keyword Search In Decentralized Storage Systems

Blockchain is one of the landing applications of decentralization.It can ensure the consistency and security of data in the decentralized network environment.However,the blockchain is more suitable for storing small objects of the“ledger”,and the storage of large objects still needs to rely on decentralized storage systems.The decentralized storage systems abandon the centralized management in centralized systems which rely on central nodes to manage data.Its bottom layer uses the P2 P network,and the status of all nodes in the network is equal.The data uploaded by each node is divided into blocks,encrypted,and encoded and then distributed directly to other nodes in the network,without going through third-party platforms.In decentralized storage systems,each object obtains a hash value after a series of hash operations,and this hash value is also the unique identifier of the object in the network.Similarly,nodes in the network can only obtain specific objects through hash values.At present,more popular centralized storage systems can provide a variety of query services according to the needs of users,such as: keyword search,range search,Top-k search,etc.Diversified query services enable them to be better used in various fields,but the development of decentralized storage systems that can only provide hashbased search is subject to certain restrictions.Centralized storage systems need to rely on different index structures to provide efficient queries.Therefore,this paper designs a global-based decentralized index for the block-access characteristics of decentralized storage systems to provide simple and efficient keyword search.In this paper,the structures of inverted index combined with B+Tree and Hash Map respectively are used as the index structures.Both index structures are kept in blocks,and both support creation,update,and query functions.Each index may not be completely saved by one node,but stored in blocks on multiple nodes,so the operations of indexes are also done based on the blocks.Because the decentralization feature will make the data of indexes decentralized,this paper also designs the version control function for the indexes,and provides suitable version merging algorithms for each index.The indexes are also decentralized,so there is no need for a central node to manage them.All nodes in the network can operate on the indexes and jointly complete the maintenance of the indexes.The experiments of indexes were conduct on the IPFS(Inter Planetary File System)cluster,the results prove that the two index structures have high availability and scalability.