Performance Evaluation Of Virtual Machines Based On D-S Evidence Theory

With the continuing development of virtualization technology,the business platform gradually migrated to the cloud resource pool.However,in the actual application process,the planning department is really concern about the large-scale business platform(huge business volume,many users,processing capacity requirements of the platform).And the source of the fear is that after the virtualization of the server,the allocation of virtual resources is crude,it will cause the performance instability.At the same time,the resource competition caused by the deployment of multiple virtual machines on the same physical host depends on the demand of applications which running on multiple virtual machines,different resource types in different times,this kind of competition makes the performance of the virtual machine uncertain.In addition,the way they use the resources and other characteristics in the same platform are not completely independent,which there is mutual influence and dependency relationship.In order to predict the uncertainty of virtual machine performance,this thesis uses D-S evidence theory to express the uncertainties that affect the performance of virtual machines.By analyzing the relationship between the various features of one certain application in the operation of a virtual machine,and the impact on the performance of this application,we can quantified the uncertainty and finally predicted the performance value of an application under certain configuration characteristics in a probability way.In general,the main work and contribution of this thesis can be summarized as follows:1.Create a virtual machine performance prediction modelD-S evidence theory not only can express the uncertainty data,but also can reason the uncertainty relationship,there is a great advantage dealing with the uncertainty problem.Therefore,this thesis uses the capacity of DS evidence theory to select five configuration indicators which may affect the performance of virtual machine,used as evidence in D-S evidence theory.And then establish a virtual machine feature configuration model based on the DS evidence theory,after that use the evidence fusion method in DS evidence theory to calculate the prediction of performance Deterministic and finally use the results of a set of benchmarks to describe it.2.Design the performance evaluation system and display the comparative experiment resultsFirst,I designed the software "based on DS evidence theory virtual machine Performance evaluation system "to show the proposed method,c#is the programming language,Visual Studio2010 Ultimate Edition is the development tool,and then combined with the functions calculation and data analysis function using interactive environment MATLAB(R2010b),parsec is the benchmark test program.Second,draw a picture of graphical user interface to show the results of comparison of experimental group and control group,in order to predict the probability section distribution of virtual machine performance.Evaluate the accuracy and necessity of the prediction that based on the accuracy of the predicted results,in order to achieve using a small amount of configured data to predict the performance of a large number of configurations.