Comparison of intelligent schemes for scheduling OSPF routing table calculation

Topology changes trigger routing protocol to undergo convergence process which prepares new shortest routes needed for packet delivery. Real-time applications (e.g. VoIP) nowadays require routing protocol to have a quick convergence time. This paper presents a new routing table calculation scheme for OSPF routing protocol to better serve real-time applications. The proposed scheme focus on speeding up OSPF networks convergence time by optimizing the scheduling of routing table calculations using computational intelligence technique. The computational intelligence technique that we use in the scheme is Feed Forward Back Propagation (BP) Neural Network. The scheme determines the suitable hold time based on three parameters: LSA-inter arrival time, the number of important control message in queue, and the computing utilization of the routers. We also provide performance comparison between our proposed scheme and another scheme which uses Generalized Regression Neural Network (GRNN). The result shows that the GRNN has higher accuracy and faster training speed compared to the BP neural network.