Research On Service Overlay Network Bandwidth Provision Based On Stochastic User Equilibrium

With the development of computer network technology, a large number of multimediaapplications emerged on the network. Traditional provision of“best effort”service network,without the protection of the quality of service, increasingly unable to meet the applicationrequirements for quality of service. Yet we expect the network to provide the“expected”end-to-end quality of service protection. Although there are a variety of QoS service programscurrently, such as integrated services and differentiated services architecture, as well asMulti-Protocol Label Switching technology, resource reservation protocol. However, thesemodels and technologies are not effectively achieved end-to-end QoS guarantees.The concept of the Service Overlay Network has been raised as a solution to theend-to-end quality of service(QoS). It is a overlay network that built on a higher level networkprotocols, operated by the third-party service provider in order to provide QoS quarantee of thevalue-added services to end users. This kind of network separated with the underlyinginfrastructure equipment, made the end-to-end quality of service possible.The healthy operation of the SON established on the basis of good business relations. Onthe one hand, SON service providers promise QoS services to attract customers to obtain income,on the other hand, SON need to buy a certain QoS guarantee bandwidth from the underlyingnetwork which incurred costs. Overall balance of these two aspects of goals, QoS-guaranteebandwidth provision model will be a very good solution, and has a wide range of applications.Domestic and foreign scholars have made a lot of research work, However, in most of theexisting research for the quantification of QoS is based on network utilization, with little regardto the influence of network traffic, and most research only for a link, with few studies of themodel.The purpose of the SON QoS guaranteed bandwidth provision is: for the specific fragment of the SON and network traffic demand, find each link needs to provide at least how muchbandwidth in order to ensure that the user’s end-to-end QoS requirements. There are twoproblems to be solved. One is how to network traffic assigned to each link. The other is for theuser’s QoS requirements need to provide the number of guaranteed bandwidth, that is themapping between the QoS requirements and bandwidth guarantee. Network assignment probleminitially started in the transportation and traffic assignment, but a series of the networkassignment methods and techniques formed by this problem can be used in many other aspects.There are two basic models: User Equilibrium and Stochastic User Equilibrium. Because of theSUE problem is more general and more realistic than UE problem, this article innovativelyapplied the SUE theory to the SON traffic assignment, in order to obtain the equilibrium linkflow distribution.This paper comprehensively consider the impact factor of the traffic characteristics, linkspeed, QoS requirements. On the basis of Stochastic User Equilibrium, use probability measuredescribes the end-to-end QoS requirements and combined with concept of the network linkutilization, explicitly determine the relationship between QoS and bandwidth guarantees. Basedon a combination of analytical analysis and simulation experiments, designs an adaptivebandwidth provision model and gives the solution of the model algorithm and simulationexample.This paper specifically discusses fixed demand of stochastic user equilibrium bandwidthprovision model and elastic demand of stochastic user equilibrium bandwidth provision model.Among the elastic demand of stochastic user equilibrium bandwidth provision model, we studythe elastic demand based on demand function and elastic demand based on network trafficprediction. Elastic demand based on demand function of stochastic user equilibrium bandwidthprovision problem can be transformed into fixed demand of stochastic user equilibriumbandwidth provision problem with network transformation method. Because of SON bandwidthprovision will not be changed within a very short time, therefore, the traffic prediction is mediumand long term prediction. The traditional timing models are difficult to fully describe themacroscopic behavior of the network traffic characteristics, they are suitable for short-termforecasts, and they will cause large error in medium and long term forecasts. In this paper, weuse BP neural network prediction to predict network traffic and the experimental results show that this approach can achieve the desired error.The bandwidth provision model in this paper, explicit the relationship between thefragment of a SON network traffic demand and the bandwidth of each link. Simulation resultsshow that the model can achieve a good mapping of user QoS requirements and network linkbandwidth. The model provides some theoretical basis for the SON network design, deployment,and decision-making, and has a very important significance for the healthy functioning of theservice overlay network.