Research On Routing And Resource Allocation Algorithm In Dynamic Optical Network Based On Prediction

With the rapid development of Internet, the scale of transport networks presents an expanding status with more complicated network forms, and the type of traffic network carried has been converted from unity to diversity. In this context, the traffic carried in optical network is vertically divided into multiple levels from the technical viewpoint, and IP/WDM network in this thesis implements the high-speed transmission and high efficiency of bandwidth utilization by combining IP technology and optical transport technology based on WDM effectively. Therefore, it has become an important development direction of the communication network. Traditional WDM network can only provide bandwidth for the smallest unit of a single wavelength. However, the bandwidth in practical communication often requires less than that of a single wavelength, and bandwidth resources are partly wasted due to this drawback. Thus, it becomes an important research subject that how to apply the grooming strategy to improve the bandwidth utilization in IP/WDM optical networks.This thesis briefly introduces the IP/WDM optical network, the optical network traffic grooming, the service prediction technology and their development status. It also gives an account of the traffic grooming, the key basic technology and the grooming solution used in routing and wavelength assignment problems. The development and application of Auto-Regressive Moving Average Model (ARMA) model prediction technology in the field of traffic forecast are then presented. Finally, this thesis puts emphasis on a new routing and resource allocation algorithm based on prediction for the dynamic traffic.Based on the characteristics of the dynamic traffic in IP/WDM optical network, a new traffic grooming algorithm, DWL-HTA-Prediction, in which historical information of traffic is used for prediction, is proposed in this thesis. When applying the grooming strategy, this algorithm adopts the history information to set up the forecast model and predict the request information including bandwidth and duration, thus choosing the optimal path and reserving resources. We select a path to carry the requested service, which has the minimum equivalent weight. The value of the minimum equivalent weight depends on the time coincidence ratio and the corresponding weight of each path Finally, through efficient grooming strategy, the algorithm could make sure that the discarded ratio keeps lowest and wavelength resources are used as much as possible.The performance evaluation of the proposed algorithm in this thesis is implemented via C programming language. The developed program is simulated in the telecom network topology that has 19 nodes and 31 bidirectional optical fiber links. The other relevant algorithms are also simulated to verify the feasibility and effectiveness of DWL-HTA-Prediction. Besides, the impact of some key parameters in the proposed algorithm on the simulation results is also studied.Finally, we summary the whole thesis and present the further research topics in the future.