Historical data learning based dynamic LSP routing for overlay IP over WDM networks

In overlay IP over WDM networks, there are only limited information exchanges between the two layers through the user network interface (UNI) [1] for service requests and responses. To enhance the overlay network performance while maintaining its simplicity, we propose for the first time to learn from the historical data of lightpath setting up costs maintained by the logical layer network operator(s), to facilitate the logical layer routing process. Using a simple cost updating strategy to alleviate the logical layer cost outdating issue, we present two dynamic LSP routing algorithms, namely Existing Link First (ELF) and Enquiries-First (ENF), for the overlay network architecture. Simulation results show that the proposed algorithms achieve much better performance than the existing algorithms with or without constraint on the number of optical ports.