As Internet traffic continues to grow unabated at an exponential rate, it is unclear whether the existing packet-routing network architecture based on electronic routers will continue to scale at the necessary pace. On the other hand, optical fiber and switching elements have demonstrated an abundance of capacity that appears to be unmatched by electronic routers. In particular, the simplicity of circuit switching makes it well suited for optical implementations. We present what we believe to be a new approach to optical networking based on a paradigm of coarse optical circuit switching by default and adaptive rerouting over circuits with spare capacity. We consider the provisioning of long-duration quasi-static optical circuits between edge routers at the boundary of the network to carry the traffic by default. When the provisioned circuit is inadequate, excess traffic demand is rerouted through circuits with spare capacity. In particular, by adaptively load balancing across circuits with spare capacity, excess traffic is routed to its final destination without the need to create circuits on the fly. Our evaluations on two separate real, large Internet service provider point-of-presence-level topologies, Abilene and GEANT, show that only a very small amount of excess traffic needs to be rerouted even during peak traffic hours when the circuit configurations are carefully chosen and that this excess traffic could always be accommodated using our adaptive rerouting approach. We also demonstrate that our adaptive load-balancing approach is robust to sudden unexpected traffic changes by demonstrating its ability to reroute traffic under a number of hot-spot scenarios.
© 2008 Optical Society of AmericaPDF Article