If we compare the total design costs given by the three metaheuristics, we conclude that there are not significant differences for the smallest networks consisting of 11 nodes. However, as the size and density of the networks increase, the differences between MSSTS and MS become significant for all the α-cuts. The hybrid metaheuristic MSSTS is better than V NS for 60 out of 234 instances, while V NS is better than MSSTS for 42 out of 234 instances. In the rest of the problems, both procedures reach the same design costs. Table VI reports the network design costs and the CPU times to get the best solution found during the search for the mixed integer linear model explained in section IV, which was obtained from the segment-path model. In this case, as in the three previous tables, both MSSTS and V NS perform better than MS, MSSTS being better than V NS 12 out of 39 and worse 5 out of 39 times.