Research On Cloud System Aging Forecast Menthon Based On Time Series

With the consistant development of cloud computing technology,cloud platform services are widely used in various fields,such as health care,public transportation,mobile communications,etc.The cloud system needs long time running,which will lead to the errors accumulation in the system and eventually cause the system performance declination or even crash.This aging phenomenon is not only reduces the reliability of the cloud platform,but also causes great harm to people’s property and life safety.For the above-mentioned problems,this thesis uses statistical analysis and deep learning methods to predict the aging of cloud server systems.By predicting the time threshold of software aging occurrence,perform software regeneration before it occurs,eliminate the impact of software aging on cloud systems,and avoid the loss caused by software aging to the greatest extent The main research contents of this thesis are as follows:(1)In order to collect data related to cloud system software aging,this thesis builds an OpenStack cloud platform based on load changes.First,the OpenStack cloud platform was built Then,design four load schemes for different components in the cloud system,conduct stress tests,and collect data.Finally,monitor the performance aging of the cloud platform during stress testing,filter out indicators related to cloud system aging,and provide a data basis for cloud system aging prediction.(2)Aiming at the problem of software aging,this thesis proposes a cloud system aging prediction method based on ARIMA.First,the time series data related to the aging of the cloud system is smoothed.Then,order the ARIMA prediction model,determine the model parameters,and verify the accuracy of the model through the residual test Finally,the ARIMA model is used to predict the historical time series of different components of the cloud system,and the system aging time threshold is obtained to provide support for determining the software regeneration time.(3)In addition,in order to predict cloud system aging more accurately,this thesis proposes a cloud system aging prediction method based on LSTM.First,normalize the time series data of cloud system aging related indicators,and divide the data into two subsets:cloud system time series training set and cloud system aging prediction test set.Then,the LSTM prediction model is established to optimize the hyperparameters in the model.Finally,train the LSTM prediction model,predict the cloud system aging related indexes after the training is completed,and verify the accuracy of the proposed prediction method through error analysis.