Abstract
As the Internet has grown in size and diversity of applications, the next generation is designed to accommodate flows that span over multiple domains with quality of service guarantees, and in particular bandwidth. In that context, a problem emerges when destinations for inter-domain traffic may be reachable through multiple egress routers. Selecting different egress routers for traffic flows can have diverse effects on network resource utilization. In this paper, we address a critical provisioning issue of how to select an egress router that satisfies the customer end-to-end bandwidth requirement while minimizing the total bandwidth consumption in the network.