Proportional QoS over WDM networks: blocking probability

The provision of scalable quality-of-service (QoS) guarantees on wavelength-division-multiplexing (WDM) networks is an important and challenging issue for the next generation Internet. One of the important performance metrics in a QoS-capable WDM network is the call blocking probability. Previously, a proportional differentiation model has been proposed as an effective method for scalable differentiated services provision. This model provides the network operators the ability of quantitatively adjusting the quality differentiation between service classes. We introduce this model into WDM networks with the aim of providing proportionally differentiated blocking probability to various traffic classes. An intentional blocking algorithm is proposed to implement this model at the wavelength level. In order to solve the link utilization degradation in this algorithm, we propose another intentional termination algorithm. Since the performance requirement from the network operator might be various, a hybrid algorithm is also given as a balance between the above two. These three algorithms are also suitable to TDM over WDM, where one connection only take part of the transmission capacity of one wavelength. Extensive simulation results demonstrate that our algorithms provide accurate and controllable differentiation on blocking probability between various traffic classes even in a bursty traffic situation. The infeasibility problem in proportional blocking probability provision is also discussed.