A Novel Forwarding Algorithm Using Prediction and Optimization Theory over Multipath Network

Abstract

In this work, a novel data forwarding algorithm based on prediction method and optimization theory is proposed for improving the quality of service (QoS) in real-time applications, in particular over the multipath network. The proposed prediction method is based on the time-series models of observed trends over all available paths at each pre-determined time. This predictive information is then used as an input for the forwarding decision. The forwarding algorithm is meant to minimize the packet loss rate by adjusting the forwarding rate of packets over each path for next prediction period by PSO (particle swarm optimization). To avoid solutions that operate near the capacity of the links and shift flows to less utilized links where they can increase more freely, the optimization problem also introduce the cost function. Numerical experiments with packet loss rate-sensitive traffic show that the optimal objective converges quickly to the optimal values. It demonstrates that running the proposed algorithm over multipath network can be more efficient and more flexible than equal cost multipath protocol.