Research On Resource Reservation Strategy And Self-adaptive Wavelength Partition In Optical Burst Switched

Large-scale using the technology of WDM has increased the transmission capacity of point-to-point, which brings the new challenges and opportunities. But the defect of electronic bottleneck not only increases the cost of system, but also constrains the throughput and flexibility of the network. And it also impedes the improvement of exchange rate. So only adoption optical switch could break those constrains. OBS (Optical Burst Switching) is one of the optional switching technologies proposed to implement in all optical networks. Based on the current technologies and cost restriction, OBS takes advantage of OPS (Optical Packet Switching) and OCS (Optical Circuit Switching) while avoiding the disadvantages of them. Thus, OBS offers a feasible and effective way for realizing all optical communication networks at present.Based on analysis of the architecture and key technologies in OBS deeply, proposed a new scheme using reinforcement learning agent to calculate the offset time. And modify the existed wavelength selection. Finally, design and realize the generally platform of OBS by OPNET. The major research results are showed as follow.(1) It is proposed a new algorithm using reinforcement learning agent to calculate the offset time. The algorithm assigns offset time flexibly using reinforcement learning scheme according to the request of network's load and QoS based on the protocol of Just-Enough-Time in buffer-less OBS network. The offset time will self-adjust according to fluctuate in the network and supporting QoS service so that it will adapt the change of network properly.(2) It is proposed a new algorithm of dynamic virtual lambda partitioning based on lambda-link. It added sorted by success rate of wavelength transmit based on dynamic virtual lambda partitioning (DVLP). The modified algorithm could updates the order of wavelengths which in the wavelength-node partition according to situation and traffic of local network, and adjust the wavelength-node partition due to change of local network dynamically. It not only reduces fragment of wavelength, but also improve utilization of wavelength. So it could be provide the better QoS. (3) Using simulation tool of OPNET designs a general network simulation platform of OBS. It includes the edge node of sending, the edge node of receiving, and the core edge. The sending edge node could self-adjust according to the change of network station, because it added reinforcement learning Agent and the receive function of feedback. Simulation results demonstrate that it reduces system's delay of end-to-end efficiently, and reduces burst's blocking probability of low priority as guarantee the blocking probability of high priority. So it improves the performance of OBS network efficiently.