Exploring Collective Behaviors with Short-Range Correlation between Routers

Existing approaches on Internet Dynamics pay more attention to the useror sessionbehaviors, but lack an overall outlook in a system level. This paper focuses on the correlation widely existing in the network and their impacts on the dynamical collective behaviors of Internet. Our simulations show that phase transition happens when the short-range interactions between two neighbor routers bring the system into a critical state. We illustrate how the Internet coordinates by selftuning and maintain this most efficient state which can be got dynamically. might uncover some nature of network and shed new light on the features of realistic IP network that have largely gone unnoticed in the past. We highlight the role of collective behaviors of interactions between two neighbor routers and identify their effects on the end-to-end connection. We contribute the fluctuations of the round trip time to the propagation of congestion routers in the network. Packet flow affects the behavior of routers, which in return regulates the flow. Even a non-correlated uniform packet flow from terminal can be modulated to show correlated fluctuations by going through the routers. In this paper, we use a simple network topology to show the phase transition exists in a real network and have strong correlation between two nodes nearby, which causes heavy-tailed RTT power spectra and long-range dependant end-to-end fluctuations. All the simulations are working on a professional network simulator, ns2, which is widely used in Internet research community. It can be found at [11]. Section 2 will show the simulation topology and results, section 3 will give a discussion by modeling the simulation and section 4 reach a conclusion. 2 Simulation and Results For the Internet’s heterogeneity, there are many types of topology existing in it. When studying the end-to-end behavior, we can simplify the core network to a single route linking several routers from the source to the destination by cutting off the side branches. To balance the router load, we design twenty router nodes, which can be seen as routers, linking in a ring. For each router node, it connects to (a) Figure 1: Power spectra of RTT with source rate 15kb/s, 20kb/s, 21kb/s respectively (a)