Proposal and Evaluation of Attractor Selection-Based Adaptive Routing in Layered Networks

To cope with ever-increasing size, complexity, and dynamics of information networks, there are emerging needs for highly robust and adaptive control mechanisms. In this paper, we take an approach to adopt a biologically-inspired algorithm to achieve robust and adaptive routing on layered networks. More specifically, we apply a nonlinear mathematical model of biological adaptation, called the attractor selection model. As a cell adaptively selects nutrients to synthesize in accordance with dynamically changing environmental nutrient conditions, in our proposal a node adaptively selects a next-hop node to forward packets in accordance with dynamically changing network conditions. Furthermore, we allow layered routing mechanisms to share their objective function and make them behave in cooperative manner to achieve higher performance in terms of speed of convergence and link utilization. Through simulation experiments, we show that adaptive routing mechanism can accomplish traffic distribution among links. Furthermore, we reveal that explicit interdependency among layers leads to lower and fairer link utilization.