Performance Evaluation And Modeling Method Research Based On IaaS Cloud Platform

With the development of big data and cloud computing techniques,more and more users build various IaaS(Infrastucture as a Service)cloud platforms.At the same time,the application scenarios of the cloud platforms are broader.For example,many colleges increase the hardware investment in the IaaS cloud platform to build computer-aided teaching rooms.Enterprises construct IaaS platforms to meet the demands for capacity of computing and storage.With the widely use of the cloud platforms,the design and implementation of the performance evaluation tools of the cloud platform have become a hot spot of academic research.At present,many researchers have designed a variety of cloud platform performance evaluation tools for different application scenarios.Through these tools,people effectively achieve the performance evaluation and management for a variety of cloud platforms.When building IaaS platforms,cloud providers need to estimate the number of virtual machines and the tasks of each virtual machine,in order to estimate the investment in the Iaa S cloud platform.However,the current cloud platform performance evaluation tools are hard to predict the maximum number of virtual machines that the cloud platform can launch.Therefore,this paper proposes a performance evaluation tool based on the IaaS cloud platforms.This evaluation tool realizes the collection and analysis of the performance data of the IaaS cloud platforms and then predicts the maximum virtual machine number of cloud platform can launch,helping universities and enterprises to budget for the IaaS cloud platforms.The configuration and the workload of the virtual machines are viewed as the key factors that affect the maximum number of virtual machines to be launched.According to the sample data of these factors obtained from the experiment,the multiple linear regression model and the support vector machine regression model are established.Then the cloud platform performance evaluation tool realized in this paper is used to test the IaaS cloud platform.Through the comparison between the fitting error and the prediction error,support vector machine(SVM)model based on the Gaussian radial basis function(RBF)is identified as a prediction model for the cloud platform performance evaluation tools.The main work of this paper is summarized as follows:(1)Based on the research and analysis of a large number of cloud platform performance evaluation tools,the characteristics and application scenarios of these performance evaluation tools are summarized.(2)When selecting the independent variables of the performance modeling,the configuration of the virtual machine in the cloud platform and the workload of the virtual machines are regarded as two key factors that affect the maximum number of virtual machines.The IaaS cloud platform performance evaluation tools collection and analysis of cloud platform performance sample data of these factors,through the comparison between different models,this paper chooses the support vector machine(SVM)model to predict the number of the maximum virtual machines.(3)The performance evaluation and prediction programs are implemented as a tool suite,including: workload generator,load module definition,performance data collector,the virtual machine configuration extractor,performance data parser,and prediction model.This performance evaluation tool can predict the maximum capacity of virtual machines which can be launched in a cloud platform.