Effective Implementation of void filling in OBS networks with service differentiation

In this paper the problem of channel reservation in Optical Burst Switching nodes is considered with the aim to reduce the scheduling processing time and, at the same time, to maintain burst loss probability as low as possible. A new implementation of the void filling problem related to a multi-class traffic environment is proposed based on the binary heap tree data structure. Fast search of the void intervals for burst scheduling is achieved by means of vector implementation of the tree that contains information about voids. The search is further optimized by reducing the search interval to obtain a good trade off between processing time and burst loss probability in relation to traffic load. Results obtained by simulation show interesting improvements of the scheduling time compared with other implementations of void filling, while, at the same time, maintaining burst loss probability low, especially for high quality traffic.