Research On The Key Technologies Of QoS-Allocation Based On Non-cooperative Game Theory Model

Nowadays, Internet users tussle with each other for their own interests. Along with the network multimedia application being widely applied and popularized, the tussle gets more and more severe. Therefore, these multimedia networking applications require better QoS-support provided by network and imminently need dynamically and efficiently QoS allocating. However, Internet cannot improve its availability to meet above requirements due to its best effort service, heterogeneity and dynamic characteristics. Research on QoS allocation support has received more attention.The research and performance evaluation of network QoS allocation technologies have been carried on in the sight of engineering aspects for a long time, such as the link utility and throughput. While putting emphasis on improving technology performance, these methods ignore the non-cooperative behavior of the end users. In essence, with comparatively high performance, the design and optimization of traditional key technologies in QoS allocation cannot rationally explain the behavior of the end users. Users are not willing to be bound in the current Internet that lacks the power of restriction. That is, we cannot make sure every user voluntarily adopts and abides by the protocols and technologies mentioned above. Some users tend to break through the constraint by all means. A typical case is adopting UDP protocol to transfer data without TCP congestion control.This will be an obstacle to the development of Internet and lead to Internet congestion collapse. Accordingly, it is necessary that new method and theory be introduced into the research of network QoS allocation technology. Game theory is surely one of the mathematic tools and provides mathematic basis for these research. In the paper game theory is applied into investigation and research on the key technologies of the flow and congestion control, routing decision and resource allocation that impenetrate the whole QoS allocation. The essence of non-cooperative behavior of end users is also analyzed comprehensively.In the part of the research of the flow and congestion control, the thesis discusses the Nash equilibrium result of flow velocity in the non-cooperative congestion game model, with which we propose Flow Control Asynchronous algorithm based on Game theory for Single Link (FCAG-SL) and Flow rate Control Synchronous algorithm based on Game theory for Single Link (FCSG-SL) that meet the needs of TCP friendly. Then a multi-follower and multi-layer Stackelberg congestion game model is presented in a creative way, which can be applied to research into the game behavior of the network end systems that have various tiers. We also design and implement Hierarchical flow Control Algorithm based on Stackelberg game (HCAS) of the various-tier end systems. At the end of this part, research on congestion game model is conducted and several effective punishment strategies are studied so as to regularize the non-cooperative behavior of end users. On the basis of these aspects, Flow Control Algorithm based on Repeated game (FCAR) is proposed.To investigate the routing decision, selfish routing game model is discussed and the reasoning of analyzing result of the routing equilibrium is carried out. The thesis also describes the traits of'atomic'routing game and'non-atomic'routing game. Then we design and implement Selfish Routing Algorithm based on non-cooperative Game theory (SRAG). Furthermore, we combined selfish routing game with congestion game creatively.As for the part of the research of resource allocation, a resource allocation game model is proposed and a resource price evaluation mechanism that based on the price contest of the end systems is designed. In addition to these, we also put forward the avail function of end system that can reflect price contest risk and propose Resource Allocation algorithm based on Non-cooperative Game theory (RANG).In the thesis, all the algorithms proposed and designed are analyzed and proved via simulation experiments. In the end, the conclusion of our research work is given and possible future research directions are pointed out.