The Obs Assembly Algorithm

Optical burst switching (OBS), which is a trade-off between the circuit switching and optical packet switching, is a promising optical switching technology. In OBS, a data burst is an intermediate switching granularity comparing to circuit switching and packet switching. Optical burst assembly is a critical technology of OBS, and will significantly affect the performance of network. The ingress router is the core of the OBS assembly. The coming IP packets are assembled into a burst and forwarded through the network as one entity according to the same destination and Quality of Service (QoS), then data burst and control packet are forwarded optical channel and O/E/O control channel respectively. The researches of foreign scholars show that network traffic streams have self-similarity (Long-Range Dependence), which has a somewhat impact on the network performance. This characteristic can result in increase of packet loss rate of small buffer system and queuing delay of big buffer system. Proper OBS assembly algorithms can decrease self-similarity of the network traffic, reduce average packet loss rate and enhance the OBS network performance. The thesis mainly studies the influence that OBS assembly algorithms have on self-similarity and OBS network performance, analyzes the factors which affect the self-similarity and performance of OBS network. In addition, we deeply discuss and validate the relationship between self-similarity of network traffic and assembly algorithms through theoretic and simulating methods. A modified united assembly algorithm with the adaptive length threshold and a united algorithm with the adaptive time threshold are put forward based on the former assembly algorithms. The two can alter the assembly parameters adaptively according to the coming network traffic. The modified united assembly algorithm with the adaptive length threshold can reduce self-similarity and average packets loss rate more than that of the former. The united assembly algorithm with the adaptive time threshold is best of all in the whole.