The simulation and analysis of phase transition in computer networks

Phase transition is one of the typical behavior patterns in computer networks. It affects the traffic management and congestion control greatly and proposes new challenges to the robust control of large-scale networks. This paper introduces a new feedback network model to investigate the phase transitions other than the cellular automata model used in the past. The experiments show that the resource competition and delay feedback in traffic flow control play important roles on the emergence of phase transitions. The transition points are also proved to be in sensitive to network sizes in the simulation. The criterion of the phase transition points and the boundaries of the link's bandwidth utilization near these points are also studied. The packet's increasing rate, decreasing rate and the switching probability between these two policies are three key factors affecting network behaviors. The network can be adjusted to just below the critical point by these three factors to get better performance and resource utilization while still keeping it stable.