Edge Aggregation. Obs Network And Shaping Algorithm

Optical Burst Switching (OBS) is considered as a promising switching scheme in optical domain, which inherits the merits of both Optical Circuit Switching and Optical Packet Switching. It takes advantage of huge bandwidth within the optical fiber and the convenience of electronic processing, so that it can overcome the disadvantages of current optical switching schemes. Although recent researches on OBS are still in theoretical level, as time goes by it would hopefully become the core technology in future generation Optical Internet.As one of the key issues in OBS, burst assembly determines how to assemble the network data packets into bursts which will be transported in the switching network. The burst assembly scheme determines the characteristics of the bursts, and would affect the network performance in many aspects. Recently, lots of researches indicate the real network traffic has self-similarity and is much burstier than traditional traffic model. So how to make the burst assembly adapt to the frequent change of traffic in order to improve the network performance is worthy to be studied.Based on the analysis of several assembly algorithms, an improved adaptive assembly scheme is proposed, which is called Improved Adaptive Assembly Mechanism with Stable Assembly Delay (IAAM-SAD). IAAM-SAD has a length threshold which can be adjusted adaptively. First it can track the change trend of the real time traffic load, and then adopts the differential adjustment strategy, makes the length threshold adapt to traffic load. The simulation under bursty self-similar traffic is made, which indicates the goal of improving network performance is achieved by IAAM-SAD. It stabilizes the assembly delay of each burst, alleviates the network performance deterioration due to the rapid change of traffic load, and avoids the shortcomings of mentioned assembly schemes.But IAAM-SAD cannot reduce packet loss rate in OBS network effectively. As OBS is a distributed network, the large packet loss rate is a critical issue. But just improving the assembly scheme is not an effective way to reduce the packet loss rate.So Traffic Prediction and Token-Based Shaping Scheme (TPT-SS) is proposed. It is an open-loop burst contention solution. TPT-SS predicts the traffic volume periodically, and then adjusts the assembly parameters and token generation rate in shaping module according to the prediction results. The bursts are scheduled onto corresponding wavelength channels by acquiring the tokens. The simulation results show that TPT-SS can smooth the bursts sequence which will be sent to switching network, so that the packet loss rate is reduced. In the simulation, the cost and setting of several key parameters in TPT-SS are also discussed.Finally, the general OBS network simulation model is introduced, on which IAAM-SAD and TPT-SS are realized, and simulation, results analysis and performance evaluation are made.