On the computational complexity and effectiveness of "N-hub shortest-path routing"

In this paper we study the computational complexity and effectiveness of a concept, we term "N-hub shortest-path routing" in IP networks. N-hub shortest-path routing allows the ingress node of a routing domain to determine up to N intermediate nodes ("hubs") through which a packet will traverse before reaching its final destination. This facilitates better utilization of the network resources, while allowing the network routers to continue to employ the simple and well-known shortest-path routing paradigm. This concept has been suggested in the past but this paper is the first to offer an in-depth investigation of it. We apply this concept to the routing problem of minimizing the maximum load in the network. We show that the resulting routing problem is a difficult (NP-complete) problem and that it is also hard to approximate. However, we propose efficient algorithms for solving this problem both in the online and the offline contexts. Our results show that N-hub shortest-path routing can increase the network utilization significantly even for N=1. Hence, this routing paradigm should be considered as a powerful mechanism for the future datagram routing in the Internet.