Research On Repairing Failure Node Considering Node Selection And Storage Node Selection For Replica Data In The Cloud Storage System

Massive data resources have been generated,which gives rise to higher requirements for the development of some emerging industries,such as big data and the Internet of things.Traditional storage systems cannot cope with massive data storage due to their limitations in cost,agility,and scalability.Distributed storage systems have become the preferred storage platform for enterprise data centers due to their flexibility,scalability,and low cost.The distributed storage system adopts the divide-and-conquer strategy to store data on multiple nodes.Meanwhile,the redundancy strategy is adopted to ensure the security and integrity of the data stored on each node.To improve the network performance and service quality of distributed storage system,this paper studies the problem of data recovery of faulty nodes and the selection of nodes for storing replica data in distributed storage systems.Two fault-tolerant mechanisms are often used to ensure the security and integrity of stored data in distributed storage systems,namely erasure coding technology and multi-copy technology.The storage overhead of erasure code technology is lower than that of multi-copy technology.However,the calculation overhead and transmission overhead of erasure codes are too large and the repair efficiency is low.Existing research mostly improves the repair efficiency of erasure codes by improving the encoding and decoding mechanism of erasure codes and optimizing the repair topology.However,the improvement of erasure code mechanism is mostly at the expense of storage space,and only consider the impact of network link bandwidth on the repair process when optimizing the repair topology,while ignoring the impact of node performance on repair efficiency.In the process of erasure code repair,selecting different participating repair nodes will also affect the repair efficiency,so it is necessary to consider the selection of participating repair nodes.In addition,after the copy data is generated by multi-copy technology,the placement of copy data will affect the system performance.Positioning redundant data on an appropriate storage node not only is conducive to decreasing the perceived access delay,but makes a great contribution to improving the fault tolerance of the storage system.This paper studies the problem of single-node failure data recovery and storage node selection for the copy data in distributed storage systems.The main innovations are summarized as follows:(1)Aiming at the low repair efficiency of erasure coding technology in cloud storage systems,this paper proposes a method for repairing failed nodes that considers node selection.Firstly,the software-defined network(SDN)technology is used to measure and obtain network status information and node load information,and the nodes that participate in the repair are determined through the node selection method based on the multi-attribute decision-making model,using the multi-attribute decision-making method to select the optimal supply data nodes and newcomer nodes.Then,the maximum spanning tree algorithm is used to find the maximum available bandwidth path between newcomer nodes,and an optimal repair tree is constructed with the newcomer node as the root and the rest participating repair nodes as leaf nodes.Simulation experiments show that the repair delay of the method proposed in this paper is significantly reduced.(2)Aiming at the problem of high repair delay caused by improper placement of replica data in cloud storage system,a replica storage node selection method based on fuzzy multi-attribute decision-making model is established to determine replica data storage nodes by considering the system network status information and node load information.Firstly,SDN technology is adopted to obtain some network link-state information such as bandwidth and delay,and the optimal transmission path can be found through the improved Open Shortest Path First(OSPF)method.Then,in accordance with the load information of node and the state of the optimal transmission path,a fuzzy multi-attribute decision mathematical model which is on the basis of grey relational degree is established.Finally,the duplicate of storage node can be determined in terms of the assessment results.Simulation experiments prove that the proposed approach in this paper effectively subserves the reduction of recovery latency.