We propose a methodology to design the link cost function and, consequently, a systematic form to design a RWA algorithm. We call this methodology link cost function design (LCFD) and it consists of four steps: The choice of the link cost function input variables, the expansion of the cost function in terms of a series, the selection of an overall network performance indicator as the optimization target, and finally, the execution of an optimization process to find the series coefficients that optimize the network performance indicator based on off-line network simulations. The optimization process is performed by a computational intelligence technique, the particle swarm optimization. The proposed methodology (LCFD) is used to design an adaptive IA-RWA algorithm, which we call Power Series Routing (PSR). The effectiveness of both methodology and IA-RWA algorithm is investigated. The PSR is compared with other algorithms found in the literature by means of computational simulations and our proposal presented lower blocking probabilities with shorter computation time. Furthermore, we investigate the sensitivity and the ability of the proposed PSR to adapt itself to topological changes in the network due to both link/node addition/failure. We also investigate the behavior of the PSR in a scenario where the traffic load distribution is randomly chosen (non-uniform traffic), and we compared it to other three routing algorithms.
Read full abstract