Research And Implementation Of Cloud Resource Monitoring And Prediction Algorithm Based On Linear Regression And Mar-Kov Chain

With the development of Cloud Computing,it provides more and more rich functions and the scale of the management of computer clusters has gradually increased,so cloud infrastructure and the deployment of efficient management is a hot point.Monitoring tools and monitoring techniques,collecting relevant information to provide administrators or users to make the appropriate decisions,can be an important role.Based on the experimental data of the OpenStack cloud environment and the analysis of the monitoring data from a commercial cloud platform,the dissertation disclosures and finds out the hidden rules and characteristics of the data.By analyzing the application of linear regression and Markov chain two prediction algorithms in predicting the application of monitoring data in cloud environment and their defects,the algorithm is used to optimize the prediction algorithm and improve the hit rate and reduce the algorithm in some overhead.The main work of the dissertation is as follows:Firstly,research cloud computing monitoring,through the analysis of history datasets to find the regulation and the features in the datasets:one day can be divided into two different time periods:busy period and smooth period.Different number of user visits that results in differences in resource consumption in different periods,which in turn affect the different characteristics of the data collected in different periods;Secondly,research on time series prediction algorithm,proposed a new algorithm model for the characteristics of data,combined with Linear Regression algorithm and Markov chain algorithm(line discrete time markov chain,called L-DTMC algorithm).What's more,according to different time periods to use different algorithms.And then decide whether to update the algorithm model and whether it needs to send update data to the server based on the Error Tolerant Degree(ETD);Finally,through the deployment of a small cloud platform,respectively,for the monitoring node and data server data consistency,the algorithm's prediction effect and the implementation of the algorithm on the impact of the three aspects of the algorithm proposed in this paper simulation test,the validity of the work of this paper is verified by analyzing the experimental results.