Research On Key Technologies Of Quality Of Service In Optical Packet Transport Networks

Quality of service is one of the core objectives of the network, but the traffic packetization, the service diversification, the network complication make it necessary to reconsider the traditional quality of service system in the transport networks. The optical packet transport network with high quality should concern about the quality of service in the whole process of delivery. Meanwhile, increasing the intelligence is also urgent for the optical packet transport network in order to ensure the optimal operation of the network and lower network operation cost all the time; in addition, related technologies also need to adapt to complex network and service requirements to meet the demands of the future multi-service transport networks.This thesis focuses on the core mechanisms and key technologies of quality of service in optical packet transport network. The main contributions of this doctoral dissertation are summarized as follows.1) To acquire the type of bottom layer services, this thesis proposes a bottom layer service identification technology based on the bit stream correlation characteristics. The bottom service type can be recognized by analyzing the received bit stream correlation features. Theoretical analysis and software simulation indicate that the method is effective in distinguishing different bottom layer service types, so that the network devices can automatically choose a proper coding/decoding scheme without manual configure.2) To identify the application type of the current traffic flows, the thesis proposes a novel online identification method using Bayesian Networks to analyze the characteristics of the head packets of the application flows. Through building up a Bayesian Networks model of packet lengths, inter-arrival times and application types, the type of an unknown flow can be identified. Using the campus network data trace to verify the proposed method, the results showed that the classification accuracy can be up to more than 95%.3) To identify the application type of the current traffic flows, the thesis proposes an online traffic classification method using estimated protocol processing time and discrete two-dimensional probability distribution. This approach builds up a series of discrete joint probability matrices of "estimated protocol processing time" and "packet length", convert the information of the unknown flow into addresses, read the corresponding units in the matrices, and then judge which application type the flow belongs to. The experiment results show that the classification accuracies of the major application are greater than 95%.4) In the aspect of QoS delivery, the thesis establishes the probabilistic graphical model to describe the interaction among these flows. Based on the dependence sources, the thesis establishes link-based graphical model and flow-based graphical model, analyzes the equivalence between these two models, and conduct the random simulation to demonstrate the existence of the network data flow dependence. The thesis also analyzes the influence of a new data flow on its neighbors. What's more, based on the establishment of the network potential energy function and the usage of analogy between the network system and the particle system, the thesis propose an algorithm to select a path for a new flow in the candidate routes set so that the interference to the existing traffic is minimized. The simulation results show that the proposed method is better than other related algorithms based on the total packet loss rate measure.