| With the explosive growth of data,there is an urgent need for a storage system that can satisfy both large storage capacity and low storage cost.In recent years,the decentralized storage system based on block chain technology has become the storage trend of data backup,which not only provides the current needs,but also guarantees the high availability and security of data.To ensure the reliability of data,Erasure Codes with high reliability and low storage overhead must adopte in decentralized storage because of the heterogeneity and instability of nodes.However,there are three problems which are low rate Reed Solomon codes of the same parameters currently applied in decentralized storage.First of all,the high repair bandwidth of RS codes.Secondly,the high overhead disk I/O results from accessing k times the help nodes to repair.Thirdly,some stable nodes' resources was wasted by the application of low rate coding with the same parameters in the uneven node environment.In this paper,three erasure codes based credibility in decentralized storage are proposed.(1)Firstly,according to the decentralized heterogeneous environment,the nodes are measured in many aspects,and a classification model is formed which divides the nodes into high trusted nodes and low trusted nodes.Secondly,based on the model,a hybrid code LRC-RS is proposed,which can reduce disk I/O,repair bandwidth and storage cost of stable nodes by making high trusted nodes use Locally Repairable Codes and low trusted nodes use low rate RS Codes.Then,the range of LRC-RS hybrid coding is analyzed in constantly fluctuating heterogeneous condition.Finally,combining theory with experimental results,the hybrid code LRC-RS can effectively reduce disk I/O and repair bandwidth with the same redundancy and specified reliability.(2)The Hitchhiker code is easy to implement under the Piggybacking structure of double substripe MDS(Maximum Distance Separable)code.However,the particularity of decentralized environment leads that the Hitchhiker code is difficult to apply.Therefore,it is the Hitchhiker code based on credibility in heterogeneous decentralized storage that is proposed.By this means,data nodes and partial parity nodes in the first substripe of Hitchhiker code are piggybacked over the remaining parity nodes in the second substripe by piggyback function to reduce the repair bandwidth of data nodes and partial parity nodes.Then let the high trusted nodes store the rest of the parity node,the low trusted node store the data node and part of the parity node,at the same time let different types of nodes use different repair strategies,so as to further reduce the repair bandwidth.Finally,the theory and experiment prove that the unclassified Hitchhiker code can reduce the repair bandwidth by about 0.5%,in addition,the credible Hitchhiker code can reduce the repair bandwidth by about 1%.(3)Combining the advantages of LRC and Hitchhiker code,a decentralized storage LRC-Hitchhiker code is proposed.Firstly,LRC divides each node into two substripes according to the Piggybacking architecture.Secondly,each set is coded according to Hitchhiker code,and the data of the first substripe is piggybacked local parity of the second substripe by piggyback function.And then let the global parity piggyback the all data in the same way.Then,the parity node is stored on the node with high trust,and the data node is stored on the node with low trust.At the same time,different repair strategies are adopted in different node.Finally,the theoretical and experimental results show that LRC-Hitchhiker can reduce the repair bandwidth by about 10% at the same disk I/O overhead. |
PDF Full Text Request