A new metaheuristic based on the Snake One algorithm is presented and it is compared with different heuristics for the NSFNET network. It is based on monitoring the congested nodes, but with sufficient coverage to satisfy the request for service. That condition does not improve essentially the blocking probability, but it does so in terms of the use of the network's resources. In the case of the blocking probability, this metaheuristic performs better in optical networks that support traffic loads up to 120 erlangs, and in the case of the use of the network it operates much better than its predecessors Snake One and Snake Two, using fewer network resources for the same traffic load.