An approximation algorithm for minimum-cost network design

This paper considers the problem of designing a minimum cost network meeting a given set of traac requirements between n sites, using one type of channels of a given capacity, with varying setup costs for diierent vertex pairs (comprised of a xed part plus a part dependent on the pair). An approximation algorithm is proposed for this problem, guaranteeing a solution whose cost is greater than the optimum by a factor of at most O(log n) (or constant in the Euclidean case). The algorithm is based on the use of a lightweight distance-preserving spanner.