Traffic Engineering Between Neighboring Domains

Network operators control the flow of traffic in today’s IP networks by manipulating the configuration of the routing protocols running on their routers. Although intradomain traffic engineering is largely isolated to individual Autonomous Systems (ASes), changes to interdomain routing policies affect the traffic load in other domains. In this paper, we argue that neighboring ASes should cooperate when changing how traffic flows between domains. Rather than proposing a new routing or signaling protocol, we argue that neighboring ASes can work in the context of the existing interdomain routing protocol— BGP (Border Gateway Protocol)—by employing network management tools that predict how configuration changes would affect the flow of traffic. Before modifying the configuration of the live routers, network operators can evaluate the effects of a proposed change and convey this information to the neighboring domain. This domain can, in turn, apply the same kind of tool to evaluate the effects of the proposed configuration change. We describe how to limit the amount of information that the neighboring ASes must exchange. This is crucial for the scalability of our scheme and for protecting the proprietary data of each domain. We also describe how a pair of ASes can control the traffic traveling between their networks without affecting how traffic flows through other domains not involved in the negotiation.