Research On Cloud Platform Software Aging Prediction Method Based On ARIMA-LSTM Hybrid Model

Cloud platform is an emerging computing paradigm.Because it can provide users with computing,storage and other services on demand,it is widely used in industry.However,during its long-term high-load operation,the system has accumulated errors and physical resources have been consumed in large quantities,which will lead to the occurrence of software aging phenomena,which will reduce the performance and reliability of the software system.In order to avoid the occurrence of these software aging phenomena,software rejuvenation technology came into being.It gradually recovers system performance by occasionally or periodically cleaning the internal aging state of the system to avoid more serious failures of the system.In the field of software aging and rejuvenation,the most important issue is accurately predict the most likely time of software aging.However,in the existing research on advanced prediction of software,since the model based on the measurement method only makes a single prediction for the linear part or the nonlinear part in the complex time series,the prediction accuracy needs to be improved.Based on the above problems,this dissertation proposes a new ARIMA-LSTM software aging prediction model,which combines the Autoregressive Integrated Moving Average model(ARIMA)and Long Short Term Memory(LSTM)model to better fit the linear and nonlinear parts of resource utilization time series.The mixed model construction process mainly includes three steps.First,the ARIMA linear model is used to filter the linear part of the original resource utilization time series,and the residual sequence containing nonlinear part and white noise is obtained,then the LSTM nonlinear model is used to filter the nonlinear part of residual.Finally,the relationship between linear part and nonlinear part is obtained by using LSTM model again,so as to obtain the final predicted value.In order to verify the effectiveness of the proposed method,this paper conducts experimental verification by using the public data set of Alibaba cloud platform,and simulates the process of software aging in the Virtual Machine Monitor(VMM)by using CloudSim simulation software to determine the software aging threshold.The experimental results show that the ARIMA-LSTM hybrid model has improved by about 50.6%,51.7%,45.8%,11.0%,13.2%,17.9%on the three prediction evaluation indexes of Root Mean Square Error(RMSE),Average Absolute Percentage Error(MAPE)and Average Absolute Error(MAE),compared with the single model ARIMA and LSTM,and the software aging threshold of VMM is 95%.Therefore,the ARIMA-LSTM hybrid model can predict the software aging of VMM in the cloud platform more accurately.When the predicted value reaches 95%,software rejuvenation operation is performed to avoid the occurrence of software aging,thus improving the availability and reliability of VMM.