Conflic Solution Of Application And Network Resource Allocation In SDN Based On Game Theory

In the network resource allocation,the application and the network optimization goals are not consistent.The optimization goal of the application is often the best user experience,and the optimization goal of the network is often load balancing,energy saving and so on.These goals are often contradictory.In addition,the application and the network have different decision variables.The network can achieve its optimization goal by adjusting the decision variables such as path and bandwidth.The application also has its own decision variables,such as distributed Web applications can control which requests each server to respond.Obviously,the decisions made by the network and the application are interdependent.But each in the decision-making to take a rational,nonmalicious attitude,are trying to pursue a win-win situation.The resource allocation model at this time is based on the framework of the game theory.It has been shown that if there is no proper coordination mechanism,the application and the network decision are alternated each time.In general,the two will not converge,the network performance will continue to oscillate,and even if the convergence,the convergence of the solution is not the optimal solution in the global sense.Based on the above situation,this thesis take“application and network cooperation resource allocation" as the core ideas,pursue solution based on the game theory.The purpose is to solve the conflict between application and network and compromise the interests of both parties.In the second chapter,study the cross-layer game problem of application and network under the scene of overlay network.Application and network Nash game,the results of oscillation do not converge,poor performance.In order to reduce the oscillation of Nash game,the author proposes a cooperative resource allocation scheme NFOR.Compared to non-cooperative Nash games,NFOR oscillations are small and application and network performance is improved.NFOR still has a certain gap between the optimal solution,and oscillation still exists,making the entire system consumption still exists.In reality,the status of networks and applications is generally unequal,the network holds more information,has the decisionmaking power of the entire network traffic,the network in the process of resource allocation in the Leadership position.Therefore,the third chapter of this thesis model the cross-level game as Stackelberg game,design SGA to solve Stackelberg equilibrium solution,propose application and network resource allocation conflict solution based on Stackelberg game,the simulation results are not oscillating,and the Stackelberg solution is a compromise.Stackelberg game is mainly from the perspective of the network to allocate resources,if the network in the Stackelberg game is willing to sacrifice its own interests appropriately,can benefit for the application.Based on these considerations,the authors creatively put forward the friendly game model friendly Stackelberg(FS),designed the friendly mechanism to the application,and studied the compromise of different friendly coefficients.Chapter 3 concludes that the Stackelberg game can solve the conflict between multiple target conflict between Overlay and network,and can also use Friendly Stackelberg to achieve different compromise effects.In addition to the above research scene,in reality,there are some typical scenarios for application and network resource allocation.So the fourth chapter studies the applicability of the Stackelberg game in the following scenarios.(1)multiple Overlay corun on top of the ISP,stackelberg game with multiple followers.(2)content service provider's server selection and ISP's Stackelberg game.(3)Incomplete information Stackelberg game: the network does not know the trasaction of the application.In the 5th chapter,the author establishes SDN experimental platform which is based on based on Floodlight and Mininet to verify the conclusion of theory simulation,including the design of experimental platform,the deployment of experimental software(sFlow,D-ITG,Floodligh,Mininet),writing modules code,platform reliability test and experimental test.Stackelberg game algorithm and the FS algorithm achieve the compromise between the network and the application under the SDN experimental platform.