Aggregation algorithms for asymmetric QoS-routing information

We consider the aggregation of link-state information for an asymmetric communication network that supports quality of service (QoS)-based routing. The common strategy used to improve routing scalability is to divide the network into multiple autonomous systems (AS). Only the aggregated link-state information is advertised among AS. Compared to the network reachability, the link-state information, such as the residual bandwidth used in computing QoS routes, changes frequently and it appears in simulations to be asymmetric. We focus on networks that employ rate-based service discipline schemes and map delay constraint into nodal rate guarantee as done with the guaranteed QoS proposed for the Internet; hence, we propose aggregation algorithms for link residual bandwidth, based on existing aggregation approaches of QoS-routing information, the full-mesh, star and spanning-tree approach. We evaluate the proposed algorithms with different traffic volumes and numbers of border nodes, based on simulations with a randomly-generated physical network. We compare these aggregation approaches in terms of their performance and scalability (the amount of link-state advertisement generated). Using our simulations, we provide heuristic solutions of achieving good performance and scalability for implementing inter-domain QoS routing. Our main results show that the spanning-tree approach with our aggregation algorithms could achieve similar performance to the full-mesh approach; both are superior to the star approach. The amount of generated advertisement traffic by the spanning-tree approach is less than that generated by the star approach or the full-mesh approach.