The Constrained Bottleneck Problem in Networks

We consider problems on networks that are captured by two performance measures. One performance measure is any general cost function of a solution; the other is a bottleneck measure that describes the worst (maximum cost) component of the solution. The paper contains algorithms to solve three problems. In one problem, we minimize the bottleneck subject to a constraint on the generalized cost. In the second problem, we minimize the generalized cost subject to a constraint on the bottleneck. In the third problem, we consider the two criteria simultaneously and find all the Pareto optimum solutions. The major result is that the introduction of the bottleneck measure changes the complexity of the original (general cost) problem by a factor which is at most linear in the number of links.