Modeling And Simulating The Router-Level Topology Of Internet

With the rapid development of Internet technology,Internet has become a symbol of the modern information society,and plays an increasingly important role in the field of human progress and economic development. There has been increasing interest in studying how to find the objective laws of the actual network through the research of the network topology model. The research on the network topology can be helpful to study the behavioral characteristics such as the spread of the virus, network congestion, network attacks and etc. The basis of research results can guide the design which has a better performance of network system. According to the research of recent years, many of the characteristics of the internet structure have been revealed, such as small world, scale-free. Specifically, the network average path is smaller and the clustering coefficient is larger, the distribution of nodes degree obeys power-law. Power-law is an important characterized of the Internet topology, and greatly promoted the research and development of the network topology.Internet topology modeling can be divided into router level topology and autonomous system level topology. However, due to the large scale of the Internet, it is difficult to know the entire Internet topology world wide. Thus, studying topology in the national ISP level is a feasible and useful way to understand Internet topology characteristic. Most of the existing models focus on AS level, and they only consider the growth of graph and connection priority without considering the evolution of node attributes and the impact of strategy employed by ISP when adding a network node or a link. Based on the analysis of the third-largest U.S. network operator Sprint RS level topology, this paper proposes a new Internet RS level topology model.From the perspective of graph theory, the development of the network topology includes adding links and nodes, removing links and nodes, reconnection. But in fact, once the link (such as optical fiber) laid, any changes will lead to enormous cost, so remove nodes and links, reconnection rarely occur in ISP topology, only adding nodes and links is concerned in this paper.In the classical models such as BA, the dynamic growth and priority connectivity for the topology are considered. However, they fail in considering the intrinsic of nodes (such as bandwidth) and their relationships in connectivity. This paper introduces the concept of node attraction, which measures the importance of nodes quantitatively. In general, the importance of the node depends on its own attributes (such as the bandwidth) and its external conditions (such as nodes in core network). The nodes that have a high node attraction value should have priority to establish high quality and high reliability connection. Some nodes which have high attraction will be more favorable to deploy new nodes or links, and they will become the core node of the network gradually. With their attraction increasing, more and more nodes will tend to connect to them, this is the reason why network shows scale-free phenomenon. This paper computes node attraction with node degree.The nodes in the topology will be restricted by a variety of physical or logical constraints, in the process of data forwarding, each node may have or use only local information of the entire network and ISP may also be restricted by technology and cost. Therefore, the establishment of connections between nodes is always conducted in the relative local area. Thus, the geographical significance of these neighboring nodes should be considered in the adding new nodes and links on the network. According to the actual situation, this paper uses weighted Euler distance to reflect the geo-spatial distance cost.In actual ISP RS level topology, because of some objective or subjective reasons, there may be a small number of nodes connecting to the ones at the outside. Network nodes establish a connection with large probability in the local world and a small probability in the global scope make the topology appears small world phenomenon.Network traffic will increase with the increasing number of end-user, in order to reduce the burden of some overload nodes and improve the utilization of the network, ISP will add links around the nodes which are overloaded. This adjusts the network topology and also enhances the network performance. Betweenness is used in this paper to measure the load of a node. During the network enhancement stage new connections will be added into the existed topology to increase the network reliability.Power-law is used to study the inherent distribution properties of the large-scale network as the Internet, to verify the correctness of the algorithm; this paper collects 1000 and 8000 as the node number. By comparing with Sprint in the CCDF, RANK, PDF power-law behavior, the modeling algorithm in this paper has an obvious power-law behavior.Most of the current network topology analysis use static network topology parameters, usually analyze the topological characteristics by calculating the clustering coefficients and average path length. However, the use of adjacency matrix eigenvalue spectrum which is also called Laplace to analyze structural characteristics of the network would have a higher accuracy, and use heat core to dynamically analyze the topology. Through the analysis of the Laplace and heat core, the model presented in this paper is demonstrated to be capable of precisely modeling the Internet router-level topology.