Research On Placement Method In VM Scheduling Of Cloud Platform

In the PaaS cloud platform,applications submitted by the user to the cloud platform usually run in virtual machines(VMs)to meet the isolation requirements.A virtual machine,namely a guest,is created,run,and managed by a physical machine(PM),namely a host,providing certain computing resources such as CPU,memory,and disks IO,disk space,network connections and more.However,as the host,the physical machine computing resources are always limited.Obviously,the lack of host machine performance will affect the performance of the client and affect the performance of the guest application.How the cloud platform perform reasonable resource scheduling can provide services for more users without affecting the performance of the virtual machine at minimizing costs then obtain more benefits,is an important issue in cloud computing and the biggest problem faced in cloud computing.In particular,Considering the possible communication between virtual machines,placing virtual machines on physical machines in different ways will directly affect performance of guest applications.Therefore,same as other computing resources,network bandwidth is also an important computing resource in the cloud platform.How to effectively allocate network resources is also an important issue of resource scheduling in the cloud platform.This paper mainly deals with the issue of how to place virtual machines on a physical machine in a cloud data center with a hybrid network topology in a PaaS cloud,which is a branch of the PaaS cloud resource scheduling problem.This paper first suggests the problem that the graph knapsack model can only apply to one single network topology and analyzes the common network topology structure of modern cloud data centers.A division method of complex network topology in cloud data center is proposed thus the graph knapsack problem with constrans of network traffic limitation can be applied to the modeling of VM placement method in a cloud data center with a mixed topology.In this paper,a deformation of the problem is used,and the corresponding integer linear programming model is given,and the Google Optimization Toolkit is used to solve it accurately.Simulation results in four different scenarios show that the model can satisfy the limit of the performance of the host and reduce the amount of communication on the physical link by placing multiple virtual machines on the same host on the same host to meet the bandwidth limitation on the physical link,so that all the application values placed on the cloud platform are maximized.However,due to the excessive restriction of network to virtual machine communication in the model,the solution time of the model is too long when the appropriate network becomes the main factor to limit the virtual machine placement.With the further expansion of the problem,the time of the model solution will not be acceptable.