On the selection of optimal diverse AS-paths for inter-domain IP/(G)MPLS tunnel provisioning

This article proposes an architecture and algorithms to select optimal diverse AS paths for end-to-end LSPs computation. The multi-domain architecture relies upon a service plane consisting of a service broker and an AS Selection Agent. Through the broker, every domain advertises transit metrics representing its transit policies (cost, routing policies) and potentially some Traffic Engineering (TE) information. The metrics are assumed to be directional, i.e. depending on the incoming and outgoing ASs. The Agent uses them to compute AS paths based on both costs and TE constraints, considering also, if needed, local policies and statistics on past transactions stored by the broker. A set of diverse AS paths can be computed, in order to proactively increase the success rate of tunnel set-up, in the case of imprecision or absence of advertised TE information (each AS path being subsequently tested), or to meet end-to-end protection requirements. If an AS path can be activated, the source router trigger the router-level inter-AS path computation along the AS path, which is accomplished by the PCE-based architecture. Within this framework, we formalize the inter-AS diverse route selection problem with directional metrics, and compare a breadth-first search heuristic with limited depth to the optimal approach. Simulations on realistic topologies prove that the heuristic scales with the number of diverse routes, and that it has an optimality gap under the 5% at least once every two times.