Joint object placement and node dimensioning for Internet content distribution

This paper studies a resource allocation problem in a graph, concerning the joint optimization of capacity allocation decisions and object placement decisions, given a single capacity constraint. This problem has applications in internet content distribution and other domains. The solution to the problem comes through a multi-commodity generalization of the single commodity k-median problem. A two-step algorithm is developed that is capable of solving the multi-commodity case optimally in polynomial time for the case of tree graphs, and approximately (within a constant factor of the optimal) in polynomial time for the case of general graphs.