Core Concepts
This paper presents new approximation algorithms for multicommodity buy-at-bulk network design and hop-constrained network design problems, resolving an open question and providing polylogarithmic approximations.
Abstract
The paper considers two-cost network design problems where edges have both a fixed cost and a length/hop constraint. It focuses on two main problems:
Multicommodity Buy-at-Bulk Network Design (MC-BaB):
The goal is to design a low-cost network to support routing demands between given source-sink pairs, where the cost of buying capacity on an edge exhibits economies of scale.
The authors obtain a new polylogarithmic approximation algorithm for the nonuniform setting via an LP-based approach, resolving an open question.
The rounding technique uses recent results on hop-constrained oblivious routing.
Hop-Constrained Network Design:
The goal is to design low-cost networks where source-sink pairs are connected by paths with few edges (hops).
The authors obtain polylogarithmic bicriteria approximation algorithms for hop-constrained Steiner forest and set connectivity problems with respect to the optimal fractional solution.
These results are obtained by leveraging a connection between buy-at-bulk and hop-constrained problems, and using hop-constrained tree embeddings.
The paper also considers fault-tolerant versions of hop-constrained network design, where the goal is to design a network that remains connected even after the failure of a bounded number of edges. The authors provide the first approximation algorithms for these problems.