Research On Prediction Method Of Cloud Cluster Job Termination Status Based On Load Characteristics

In recent years,the Internet has developed rapidly.We have entered the era of big data.With the development of cloud computing technology,cloud platform has become a basic platform.Research on cloud platform resource management has become a hot spot.How to take advantage of the characteristics of the cloud platform load data,improve the resource utilization of the cloud platform,and provide users with better service,has become an important issue for cloud platform resource management.This thesis focuses on one of the core issues of cloud computing technology: cloud platform scheduling management,the purpose is to solve the problem of forecasting the termination status of cloud platform scheduling management,and the log data published by the large Google cloud platform will be used as the experimental data set.For the attribute analysis of the data set and the job termination status,the appropriate load feature vector samples are selected to prepare for the model input.After analyzing and comparing various aspects of the current classification prediction model(test time,prediction accuracy and scope of application),the extreme learning machine(ELM),support vector machine,online sequential extreme learning machine(OS-ELM),multilayer limit were selected.The five better models of learning machine(ML-ELM)and multi-layer online sequential extreme learning machine(ML-OSELM)predict the termination status of the cloud platform.The results show that the models are similar in terms of test time;In terms of test accuracy,the test accuracy of multiple hidden layer neural network models is higher than that of the single hidden layer network model.The multi-layer extreme learning machine model(ML-ELM)has the highest test accuracy of 92.67%.The test accuracy of the multi-layer online sequential extreme learning machine model(ML-OSELM)is similar to that of the multi-layer selfencoding extreme learning machine model(ML-ELM),which is 92.45%,and the MLOSELM model can effectively solve the cloud platform.Because the amount of data is increasing and it is limited by memory,the ML-OSELM model is the optimal model in the above model,and is more suitable for predicting the termination status of jobs in large-scale cloud platforms.According to the experimental results,the multi-layer online sequential extreme learning machine classification prediction model(ML-OSELM)can reduce the memory occupancy rate in the cloud platform,reduce the service cost and improve the service quality of the cloud platform,and terminate the operation status in the cloud computing scheduling management.Forecasting problems have an important role to play.