Recovery techniques in optical burst switching networks

Abstract

In this thesis, we study recovery scheme for OBS networks and evaluate performance of recovery scheme through simulation using NS-2. The main objectives of this thesis are follows. First, we redesign the recovery procedures considering characteristics of OBS networks. General recovery schemes cannot be applied for OBS network due to the offset time which is time difference between control packet and data burst. As the number of hops is increased than as expected at the ingress node, the offset time becomes getting shorter than planned, and finally the data burst is dropped. So, we have to redesign all recovery procedures considering offset time effects. Second, we implement all recovery techniques using NS-2 simulator. We have implemented recovery techniques such as restoration, pre-planned restoration and protection using well-known network simulator which is discrete-event-driven simulation. We also implement detailed recovery techniques such as path, link, and subpath. Although ns-2 provides substantial support for simulating IP networks, it lacks the necessary modules for simulating OBS recovery capabilities. So, we have to implement additional modules to simulate recovery techniques in OBS networks. Third, we propose new enhanced restoration techniques. Usually, restoration techniques have relatively longer recovery time than protection techniques. It is originated from the procedures of backup path reservation. But, because control packet reserve bandwidth for data burst, the backup path reservation procedure can be eliminated. Using this fact, we can dramatically improve the recovery efficiency in terms of recovery time and blocking probability. We then present performance of all recovery techniques using well-known performance metrics. By studying these results, we can understand the performance of each recovery techniques. The simulation results can also verify that the enhanced restoration scheme has better performance than the other recov...