A Research On Several Key Problems In Improving Network QoS Based On Traffic Measurement

With the rapid development of the Internet, a wide variety of different user applications are deployed. Especially with the advent of multimedia services, today's Internet evolves into a global commercial infrastructure that is expected to support a plethora of new applications such as IP telephony, video-conferencing, audio and video streaming, and distributed games. Unlike previous applications (e.g. file transfer), these new ones have much stricter timeliness and predictability requirements. In consequence, there is an urgent need to improve the network QoS(Quality of Service). However, the current Internet adopts the "best-effort" service model, which cannot make any explicit commitment as to Bandwidth, Latency, Jitter or any other service quality. Meanwhile, the sharp advance of network applications has dramatically increased the traffic flow of the Internet, and brought highter pressure to the network. Consequently, the manager of the network should make good use of the limited network resource to satisfy the rapidly rising bandwidth demand. Although the network bandwidth and transmission speed have greatly increased, the data traffic is increased even faster. As a result, the bandwidth and the transmission speed are still the bottleneck problems in the network.In order to make the current Internet provide better QoS guarantees, many works have been reported in the literature, and many schemes are proposed at different layers of the network. Since the user acquired QoS and the service capability provided by operaters are all obtained from the analysis of the network traffic flow, the research on the characteristics of the network traffic flow is a key problem for improving QoS. Thus, the author of this dissertation has dedicated herself to this area, and carried out the following study:1) Study on IP network traffic measurement: Network traffic measurement is an effective way to research on the network behavior, at the same time, understanding network behavior is very important for gaining a better management of network resource and quality guarantee of services. The architecture and realization of the traffic measurement system for IP network is analyzed in the thesis. The characteristics of IP network traffic flow based on Traffic Monitor and Analyzer system are also discussed, many useful results such as protocol distribution, packet length distribution and traffic quantities and derections are obtained, which provide a meaningful reference for network planning and technique research.2) Improving network QoS in BE(Best Effort) Model: In order to guarantee the bandwidth fairness, the banwidth of the service such as P2P, which has high throughput flows, needs to be restricted. We deployed queue management in the service that has the quality demand mainly on bandwidth, and proposed a new algorithm named Bandwidth-Fair RED (BF-RED). The algorithm can protect from high-bandwidth flows to consume a lot of bandwidth in some network environment, and guarantee the bandwidth fairness when congestion occoures, so that guarantee the QoS of the network.3) Improving network QoS in IS(IntServ) Model: The core of IS Model is RSVP(Resource ReSerVation Protocol). RSVP is a unidirectional resource reservation protocol, it can only reserve resources for the TCP data flow along the forward direction path, and cannot provide any protection to the ACK packets against congestion in reverse direction. In order to improve TCP performance over RSVP, we proposed a new protocol called ABi-RSVP(Asymmetry Bi-directional RSVP). ABi-RSVP reserves the backward resource by taking account of the ACK packets size, therefore improves the throughput of TCP flows and enhances the efficiency of the network.4) Improving network QoS in DS(DiffServ) Model: The DiffServ architecture does not specify any priority scheme between Assured Forwarding (AF) out-of-profile packets and Best Effort (BE) packets. In general, the AF out-of-profile packets have the priority when compete with BE packets, and consequently affected the fairness of the network. Aiming to solve this problem, we proposed a simple core-stateless queue management scheme TD&BF(traffic differentiation and bandwidth fairness). TD&BF not only guarantees the committed rates of AF flows but also fairly allocates the extra bandwidth among the BE and AF flows, so that guarantees the QoS of the network.5) Web2.0 traffic measurement and analysis: Web2.0 has becoming the new tendency of the Internet development. We investigated the architecture and the advantages of Web2.0 under Ajax technology, and compared the traffic characteristics of Web2.0 with that of the traditional one. Furthermore, the traffic characteristics of the typical Web2.0 service are analyzed.