A dynamic traffic adjustment approach to optimizing performance in multipath network serving heterogeneous flows

In this work, we study the traffic adjustment over multipath network in the presence of both inelastic and elastic traffic flows. The characteristics of these two types of traffic differ significantly. Hence, earlier approaches that focus on homogeneous scenarios with a single traffic type are not directly applicable. We formulate a new traffic adjustment problem based on time series prediction and HPSO (hybrid particle swarm optimization) that incorporates the performance requirements of inelastic and elastic traffic flows. Simulations results show that our adjustment approach not only achieve the maximization utility gained by elastic traffic but also satisfy the reliability demand of inelastic traffic. Our optimal algorithm is extended to provide better packet loss rate performance for the inelastic traffic.