| With the fast development of network technologies and the increasing scale of network, traditonal Internet with best-effort service can not meet the need of multimedia and non-traditional applications. To afford Quality of Service (QoS) according to user's requirement becomes a general request in the high-speed network as well as an active research field in computer networks.Congestion control and QoS of network are two dense interdependent subjects for research. As an important approach to prevent and solve congestion of network, congestion control can not only improve the performance and efficiency of the network, but also become the precondition for realizing network QoS. A network with frequent congestion can not guarantee good service quality.MPLS put forward by IETF is a prospective technology, which can realize QoS for the next generation backbone network. It can afford effective assurance for QoS since its ability of traffic engineering. But MPLS does not have an effective mechanism to guarantee the fairness between responsive flows and unresponsive ones, and the unfair competition for network resources between them is a great obstacle for QoS implementation.In this paper, the overview of congestion control technology at present is reviewed at first, then the deep discussion on TCP congestion control based on window and IP congestion control in router is given. The QoS architectures issued by IETF such as IntServ, DiffServ and MPLS are introduced afterwards. Based on emphasizing the important role of fairness in congestion control and comparing several kinds of AQM algorithms for fairness, a fairness-based MPLS mechanism called f-MPLS is proposed, which resolves the unfairness issue in the MPLS domain by integrating an AQM algorithm CSFQ with the MPLS model. After analyzing the principle of the mechanism and describing the procedure of it, some simulations of f-MPLS are presented and the results show it's validity and efficiency in a certain extent.
|
PDF Full Text Request