A parallel branch-and-bound method for a network design problem

A general parallel branch-and-bound system has been developed on a Sequent Balance. This paper reports its application to a non-convex network design problem, namely that of capacity determination for a backbone computer network. The results of some numerical experiments are given. From this experience it is concluded that the use of parallel branch-and-bound, particularly in conjunction with a genetic algorithm optimiser, can be both effective and profitable for non-convex network design problems.