| With the rapid development of information network, the voice, data and vide applications will be converged on a broad band IP network. But, the conventional IP networks have no explicit Quality of Service(QoS) support and only provide"best effort"services for all applications. These does not encourage mass adoption of IP networks as transport mechanisms for real-time and mission-critical applications and have brought many difficulties for the evolution and actualization of the multi-services networks. As an important IP QoS architecture, the Diffserv(Differentiated Services)techniques have good expansibility and suited to added to the network that may enforce quality of service. Now, InternetII has also set up Qbone to test the DiffServ techniques in the NGN (Next Generation Network).The thesis mainly researches how to provision DiffServ for the real-time applications with the QoS requirements and to save the WAN link capacity which determines the total costs for the enterprise. In the DiffServ network, we put VoIP traffic in one service class, video traffic with real-time requirement in another and web traffic without QoS requirement in a third service class. Each class of traffic requires a different level of QoS guarantee. For VoIP the primary QoS requirements are delay, jitter and loss; for real-time video end-to-end delay.Our simulation study for the DiffServ techniques is made in the multi-services intranet environment with NS-2 network simulator used widely by network researches. The thesis provides an overview of NS and a thorough examination of the modules added to the NS DiffServ implementation. Meanwhile, the study also made some extensions and modifications in the implementation of DiffServ architecture to meet the study requirements. In the simulation study, we have analyzed the behavior and source model of some familiar traffic, such as VoIP, Video, WWW, FTP, Telnet and customized routines for collecting performance statistics.Our goal is to measure with simulations the benefit of integration and differentiation of services. Therefore we try to answer, which of the following three cases requires lower capacity.1.Keeping VoIP, video, and best-effort data on dedicated IP networks.2.Integrating VoIP, video, and best-effort data in a conventional IP network without DiffServ support.3.Integrating VoIP, video, and best-effort data in a network with DiffServ support for the three classes of traffic (EF-VoIP, AF-Video, BE-WWW).The study found that for the enterprise intranet, an integrated network without DiffServ support could not differentiate services. Therefore we have to considerable over-provision such that all data traffic meets the QoS requirement for real-time applications. In addition, we observe significant capacity savings in the DiffServ case; Compared to the dedicated networks case, the savings in capacity are approximately 40%. At the same time, the QoS... |
