Simulation based modeling of optical burst switching networks

Optical Burst Switching (OBS) has been suggested as an intermediate step towards full Optical Packet Switching (OPS). There are considerations over the algorithms/protocols to be used. For example, different algorithms have been proposed for assembling bursts, for selecting the wavelength channel to transmit the burst on (scheduling) and for Class of Service (CoS) provision.

The first aim of the research discussed was to develop a robust simulation model for OBS which could be used in the future for developing and investigating different algorithms involved in OBS networks. We succeeded in developing a unique OBS model in OPNET Modeler that includes all stages of an OBS network. These stages include burst assembly, Class of Service and scheduling algorithms. The model is targeted at networks that may have near-term implementation potential (without wavelength conversion and optical buffering, although the model is capable of modelling such processes). The way the model was developed has made it possible for future researchers to add any burst assembly, Class of Service and scheduling algorithms and use them to investigate the performance of burst dropping probability in an OBS network using the Just Enough Time reservation technique.

The second aim of the research was to develop a new optimization algorithm for the simple scheduling schemes proposed for wavelength allocation in optical burst switching (OBS) networks. The operation of this algorithm is compared to standard first-fit and random scheduling schemes with different combinations of burst assembly algorithms. Lower burst dropping probability in comparison to other scheduling schemes is shown for our new optimized scheduling scheme.

The final aim of the research was to demonstrate for the first time the effect on burst dropping probability when combining different algorithms of all stages. It has been proved that the performance of an OBS network has a more complex interdependence between the choice of assembly algorithms and scheduling algorithms, suggesting that the choice of scheduling and assembly algorithms must be considered together.