All-optical networks require end-to-end lightpaths to be established for traffic to flow. Given that there are no wavelength converters present in the network, each lightpath only accommodate a single wavelength. It is shown that as the lightpaths traverse more hops, the blocking probability increases.This is causes the Fairness Problem. We introduce the Traffic Classification and Service Method (ClaServ), which optimizes the Fairness Problem, as well as reduce the traffic blocking probability when the networks require lower blocking probability. The combination of the Waveband Access Range (WAR) and the Waveband Reservation (WRsv) methods changes the traffic's distribution among the wavebands on each link of the path to control the degree of the interference among the classified traffic. Under certain traffic load, by setting the range of accessible wavebands and reserved wavebands for classified traffic, the network can achieve both the fairness and low blocking probability for all type of traffic. The simulation results show that for a 4x4 Mesh-Torus network the ClaServ method can greatly reduce the blocking probability for longer lightpaths by a factor of 100. It is also described how the ClaServ method can easily be implemented into a distributed signaling protocol.