Research Of Routing Protocols For The Large Scale And Multi-operation Characters Of NGI

With the rapid development of past years, the Internet has evolved into a large-scale integrated network with various data traffic such as Video, voice etc instead of homogenous data traffic. It is prevalent and challenging to provide QoS to meet all kinds of application requirement in the large-scale network. However, the current Internet architecture offers mainly a best-effort service and does not meet the requirements of integrated services networks designed to carry heterogeneous data traffic. It is a hot topic to support routing problem in large scale networks for different applications.In a long period time, the Internet needs providing two services: best-effort service and multiple constraint QoS service. Therefore, the routing model has to support these kinds of service simultaneous. To support both services, there are some questions to be solved:First, the introduction of QoS will have an impact on the best-effort services, which furthermore affects the work of regular routing protocols. Under this situation, the efficiency, flexibleness and scalability of best effort routing is very important.Second, QoS services not only need to satisfy various QoS requirements, but also keep practical.To solve the routing problem efficiently, this thesis is about the research in routing models and protocols which can support the large scale character and the multi-operation character. The performance evaluation are achieved by simulations. Through the comparison, the Network Tree routing model which designed in this thesis has the following advantages:For best effort service, the routing table is set up by the access probability, which enhances the utilization of the routing information. Also, since the routing table scale is reduced and the cost of the setup and maintenance of routing table is low.For QoS services, this thesis designs a QoS control algorithm which based on the Network Tree model. It includes admission control and fair queuing algorithm. In this schema, only edge routers maintain the flow state while core routers just check its output rate to perform local admission control. A lightweight signaling is adopted in the domain to simplify the complexity, which enhances the scalability greatly. Furthermore, it can provide fair bandwidth allocation.