The Internet Protocol (IP) has emerged as the dominant technology for determining how data is routed across the Internet. Because IP flows are defined essentially in terms of origin-destination (O-D) pairs, we represent IP traffic engineering as a multi-commodity flow problem in which each O-D pair is treated as a separate commodity. We account for the diversity in IP routing by modeling opposite extremes of traffic engineering: "naive" traffic engineering where the IP routes data between any two users using only the shortest path between them, and "best case" traffic engineering where IP has the flexibility to route data using multiple paths in the network regardless of their length. We develop linear programming formulations that identify the maximum data flow for an IP network that satisfies proportionality constraints for traffic demand for each case of traffic engineering, and we also determine the optimal interdiction of those flows that reduces that maximum flow in the worst possible way.

Approved for public release; distribution is unlimited.

http://archive.org/details/anttackerdefende109454288

US Navy (USN) author.