Performance Evaluation Of IaaS Cloud Based On PEPA

Infrastructure as a Service(IaaS)cloud integrates computation,storage,and other resources into virtual machines which are provided to users via the internet.It improves resource utilization and reduces costs.Performance Evaluation of IaaS cloud helps cloud providers to make adjustments to the platform,and as a result the quality of service(QoS)can be improved.Compared with actual measurements,performance evaluation based on analytical models requires little resource and has higher flexibility.As a stochastic modeling method,Performance Evaluation Process Algebra(PEPA)has the characteristic of combination.It represents the model of a large-scale system as the interaction of multiple models of subsystems,which makes it easier to evaluate IaaS cloud.The main research contents are listed as follows:(1)PEPA models of IaaS cloud are built.IaaS cloud is mainly composed of two parts.One is hardware resource and the other one is IaaS cloud management system.This paper constructs a PEPA model of IaaS cloud management system and a PEPA model of IaaS cloud hardware resource,and gives a method which can calculate model parameters through real data.In order to make PEPA models of IaaS cloud more accurate,this paper introduces transient actions into them.(2)Solutions to extract performance metrics from PEPA models of IaaS cloud are proposed.This paper analyzes the impacts of transient actions on some existing methods in PEPA,and makes adjustments to the existing solutions to utilization and throughput.Because existing solutions to response time can not meet requirements,a new solution to response time is proposed.This method can directly obtain the probability distribution function(PDF)of response time of a service by specifying two actions.It defines service flow as the necessary execution process to complete the service.The whole operation is divided into two steps.Firstly,extraction algorithm generates service flows automatically.Then,PDF of the response time is derived based on service flows.Because the simulation solution consumes a large amount of time and computation to obtain the PDF of response time,the approximate solution has also been proposed.The approximate solution can quickly solve the PDF of response time based on Fluid Flow.(3)Performance evaluation of IaaS cloud is performed.This paper firstly selects suitable model parameters according to the actual requirements.Then,performance metrics are extracted,and they are used to determine the relevant characteristics of IaaS cloud by comparisons and analyses.This paper focuses on response time of IaaS cloud management system and utilization,hardware load,.etc of IaaS cloud hardware resource.The final result of performance evaluations reflects that characteristics of IaaS cloud can be extracted from PEPA models,and these characteristics can be used as a reference for the measurement of QoS,resource adjustment,.etc.