Research On Replica Management Strategy In Heterogeneous Cluster

Today is the new era of the Internet,with rapid development of various high-tech,including e-commerce,self-media,news media,social platforms,etc.,have reached a new level.With the rapid development of computer technology and the emergence of 5G networks,the number of Internet users has increased dramatically,and data files and multimedia materials have increased dramatically.How to ensure data security,efficient storage,fast acquisition,processing,and improve the overall performance of cloud computing has become a hot topic in today's research.Cloud storage solves the storage problem of massive data and ensures data reliability.In order to improve the efficiency of data access in the network and enhance the load balance of nodes in the file system,the paper studies copy management strategies for these problems.The main work of the paper is as follows:(1)Based on the cuckoo algorithm to optimize the grey Markov model,a file popularity prediction model is established.The collected historical file access information is used to form the original prediction sequence,the unbiased gray model is used to predict the file access heat at the next moment,and the Markov model is optimized by the cuckoo search algorithm to correct the prediction value error,and the prediction sequence is performed using the metabolic idea Update.The prediction model has an accuracy of 96.92%,which is the most accurate compared with the neural network and the original gray model.(2)Adjust the copy factor based on file popularity and availability.A copy factor adjustment model is established based on the file popularity value to be adjusted and the average file popularity value in the cluster,a copy adjustment model based on file availability is established according to node availability,and a copy factor adjustment model is obtained by comprehensive consideration.For files with high popularity,increase the file copy factor to deal with the access delay caused by the sudden increase in access;reduce the copy factor for files with low access,but the minimum should not be less than two.Follow the security of the multiple copy strategy consider.(3)Replica placement node evaluation and selection strategy design.First,select the three factors of disk space load capacity,CPU load capacity,memory load capacity and node efficiency to establish a node evaluation model,use the analytic hierarchy process to determine the importance of the evaluation factors,and calculate the node evaluation value.According to the node selection strategy,the node with the highest evaluation value is selected as the copy placement node.The paper aims to improve the load balance and computing efficiency of the nodes in the system for copy placement.The experiment uses the CloudSim simulation platform to verify the copy placement strategy.The standard deviation of the storage space utilization rate of each node in the default method is 0.206,and the tomographic analysis is optimized.The standard deviation of the factor evaluation algorithm is 0.058.The average execution time of the copy placement strategy of the default,custom weight,and tomographic analysis optimization multi-factor evaluation algorithm is 81.16s,78.88s,and76.24s.The time growth rate of the three cases is 3.9%,3.1%,2.4%,When the analytic hierarchy process is adopted,the time increase is the smallest as the number of jobs increases.