Predictive Analysis,Modeling And Simulation Of Performance For WEB Applications

The demand for cloud computing services is growing rapidly,which has become a new trend in Internet computing.The cloud can provide flexible resource supply and pricing,so many Web applications have been migrated to the cloud.These application systems are usually network-intensive service systems,resulting in serious consumption of bandwidth resources,and due to cost control,bandwidth capacity is limited,so the increase or fluctuation of the load makes bandwidth resources often become a bottleneck.Due to factors such as network congestion and retransmission mechanism,network transmission has a significant impact on the service response time,and the service response is natural to cause a delay.Therefore,it is particularly important to predict the network performance of Web applications,which is a challenging problem to be solved urgently.At present,there are many research methods for Web application performance,but most of them are about CPU,memory,and other computing resources,but there are few studies on bandwidth resources,and the modeling process of the existing technical solutions is more complex and expensive.This paper makes a predictive analysis and modeling research on the performance of Web applications,with more emphasis on bandwidth resource consumption and Qo S.We propose an analysis method based on log mining,which realizes the automatic mining of load parameters,adopts the idea of time slicing and service grouping,and parallelizes the serial request path,thus simplifying the load model.We use three different modeling methods to predict the change of bandwidth consumption and response time when the load intensity changes,and estimate the service capacity of the server under the condition of limited bandwidth,as well as the impact of different bandwidth configuration on the Qo S,all these provide decision support for bandwidth resource management.The main research contents of this paper are as follows:(1)Through the method of analytical modeling,we study the prediction of bandwidth consumption and response time.We propose a method based on log mining to predict the bandwidth consumption and Qo S of Web services,which provides decision support for bandwidth resource management.Based on regression analysis and function fitting,this method can estimate not only the service capacity of the server under the condition of limited bandwidth,but also the impact of different bandwidth configuration on the Qo S.In this paper,we use open data sets and benchmark experiments to evaluate the effectiveness and accuracy of this method.(2)Through the modeling method based on machine learning,we study the prediction of bandwidth consumption and response time.We propose a method based on Bayesian reasoning to predict bandwidth consumption and Qo S.We use log mining to obtain feature variables,use step-by-step discriminant selection to construct a Bayesian model,and use Bayesian discriminant and Bayesian network to predict bandwidth consumption and response time.In this paper,we use a real Web system as a case to evaluate the effectiveness and accuracy of this method.Compared with other methods,this method can predict the performance of Web applications more accurately.(3)Through the method of modeling and simulation,we predict the bandwidth consumption and response time.We propose a method based on network modeling and simulation to predict bandwidth consumption and response time.Our simulation model can not only simulate the real user behavior,but also accurately predict the bandwidth consumption and response time under different load intensity,and simulate the influence of geographical factors on the response time.It can also predict the changing trend of bandwidth consumption and response time in the scenario of bandwidth resource reallocation.The significance of this paper is to predict the changing trend of network performance under the change of load or system configuration through predictive analysis,modeling and simulation for Web application performance,and to predict the possible resource consumption in the future,to realize the proactive planning and management of computing resources,which is also of great significance in performance optimization for Web application.