Study On The Key Techniques For The Edge Nodes In The Optical Burst Switching Network

Optical Burst Switching (OBS) is a promising solution for supporting high-speed bursty data traffic over WDM network, which combines the benefits of optical circuit switching and optical packet switching. Compared with optical circuit switching, OBS supports bursty traffic more efficiently by adopting one-way reservation protocol. On the other hand, OBS can be implemented easily on the basis of the current optical technologies compared with optical packet switching.As edge nodes play a very important role in the OBS networks, this thesis discusses some key techniques regarding them. The major contributions of this thesis are as follows.1. Under the Poisson input traffic assumption, the distribution of the data burst length and the burst inter-arrival time are studied in case of that the time-based and threshold-based burst assembly algorithm are used respectively at edge nodes.We demonstrate that when the IP packets arriving to the assembler is a Poisson flow and the packet size is exponentially distributed, the data burst generated can be approximated as a Poisson flow too, but the burst length distribution is not exponential and it has a close relation to the assembly algorithm used. On this basis of that, further study for the delay performance of the edge node has been also carried out. Utilizing the obtained analysis results, we propose a decision mechanism to determine the minimum boundary value of the burst assembly parameter for satisfying the switching system design requirements at core nodes and the QoS requirements of import traffic.2. In order to solve the problem of burst padding in OBS networks with light-load, we introduce a novel burst assembly algorithm, Burst Padding Reduction Assembly algorithm (BPRA), to minimize the padding overhead while maintaining the end-to-end delay requirement of IP packets. The algorithm determines the burst generating time according to the burst length and uses the offset-time to prolong the assembly time. We also utilize the numerical analysis and simulation to show the effectiveness of the algorithm.3. We present an analytical model to analyze the impact of the burst convergence mechanism on the self-similarity of traffic. Measuring the traffic volume in bytes and packets/bursts respectively, we demonstrate that only in case of the timed-based algorithm is used, the self-similarity of the assembled burst counted in bursts is reduced.