The prediction-based routing in optical transport networks

In optical networks, the traditional routing problem is generally decoupled into two subproblems, the route selection and the wavelength assignment. Usual RWA (Routing and Wavelength Assignment) algorithms take the lightpath decision based on the network state information existing in those nodes caring about it. Unfortunately, this information might not be accurate enough so that the routing decisions could be incorrectly performed hence driving to a significant connection blocking increment. The mechanism proposed in this paper aims to optimize the network performance while reducing the signalling overhead required to keep updated network state information on all the network nodes. The novel idea introduced in the Prediction Based Routing allows lightpaths to be computed not according to the potentially inaccurate network state information but according to a prediction scheme. In this way, flooding update messages is not needed so that the signalling overhead is significantly reduced. After analyzing the PBR behaviour, we conclude that the PBR performs better than usual RWA algorithms (such as the Shortest-Path/Least-Loaded) in different scenarios of traffic loads and available resources.