Dynamic Allocation Strategy Of Virtual Machine Bandwidth Based On Time Slice Prediction In Cloud Computing Environment

With the increasing demand for cloud computing services,limited bandwidth capacity has become a key bottleneck for future cloud growth.At present,the optimization problem of bandwidth allocation in cloud data centers needs to be solved urgently.Various strategies for optimal allocation of bandwidth in cloud data centers have emerged.Due to the complexity of cloud data center network topology,many bandwidth allocation strategies still have deficiencies.For example,at the virtual machine level to a certain extent,the waste of bandwidth resources can be reduced by bandwidth guarantee,but it can't completely take full advantage of bandwidth in time and space.the utilization of bandwidth resources was improved by the predictable bandwidth allocation in time and space,but it focuses on the entire cloud data center or the application itself and ignores the bandwidth allocation of virtual machines on the access layer host server,and in the virtual machine When a traffic burst occurs,the bandwidth of the virtual machine cannot be adjusted in time.This thesis focuses on the bandwidth allocation problem of virtual machines on the access layer host server,and proposes a strategy for dynamically adjusting bandwidth of virtual machines based on time slice prediction.The ideas of minimum bandwidth guarantee and predictable bandwidth allocation was combined in this strategy,adaptive time-based on-chip prediction was adopted,and historical bandwidth at this moment was used as a reference to adjust the virtual machine bandwidth.According to the above theory,the dynamic bandwidth adjustment strategy of the virtual machine based on time slice prediction is implemented based on the Open Stack platform.The research content of this thesis mainly has the following aspects:(1)Referring to the schemes of Oktopus,Gatekeeper,and Cicada,based on the Hose model,a dynamic bandwidth adjustment strategy based on time slice prediction is proposed.Through experiments,linear prediction,polynomial prediction and other methods are verified,and linear prediction methods are used to predict the flow trend of virtual machines.At the same time,it draws lessons from the sliding window mechanism in the flow control technology,and ultimately finds a way to adjust the bandwidth of the virtual machine in an adaptive time slot according to the change in the flow rate of the virtual machine,so as to ensure that the proposed strategy can quickly adapt to the flow change of the virtual machine.In addition,the historical bandwidth at the corresponding time is used as a reference to further adjust the prediction bandwidth of the virtual machine.(2)Researching cloud computing infrastructure,platform features,and operating models.Among them,taking Open Stack,a popular open source cloud computing platform,as the starting point,we focused on the Open Stack infrastructu re and its core components,learned the secondary development of Open Stack components,and mastered the traffic collection mechanism and monitoring principles of Ceilometer components.The strategy lays the foundation for the implementation on Open Stack.(3)Researching the Docker container technology,and integrating it with the bandwidth allocation strategy proposed in this thesis,designing the Intelligent Transportation Cloud System(ITS)based on the Docker container technology to solve the problem of network traffic congestion in the traditional intelligent traffic cloud convergence layer and software as a service(Saa S)layer It also solves the problems such as the low utilization of the overall resources of the transportation cloud,the inability of integration and sharing of traffic data,complex application management,and migration difficulties.