Link dimensioning in WDM networks with dynamic traffic grooming

This paper presents an analytical model to determine the number of required wavelengths to meet a specified grade of service (GoS) in WDM networks with dynamic grooming capabilities. The GoS is defined in terms of overall call blocking probability in the system. Traffic consists of two rate streams, low-rate and high-rate. For each traffic stream, call interarrival and holding times are exponentially distributed. Two wavelength assignment strategies, random and first-fit, are considered to accommodate the calls in the network, in the case of random assignment, the system can be formulated as a Markov process from which a product form solution can be approximated the overall call blocking probabilities. Numerical and simulation results show that in low traffic regime this formula is a lower bound for the random assignment and an upper bound for the first-fit assignment. Numerically, we show that the product form formula can be used to dimension the network, when implementing the first-fit assignment strategy.