Analyzing and modeling router-level internet topology and application to routing control

摘要

Measurement studies on the Internet topology show that connectivities of nodes exhibit power-law attribute, but it is apparent that only the degree distribution does not determine the network structure, and especially true when we study the network-related control like routing control. In this paper, we first reveal structures of the router-level topologies using the working ISP networks, which clearly indicates ISP topologies are highly clustered; a node connects two or more nodes that also connected each other, while not in the existing modeling approaches. Based on this observation, we develop a new realistic modeling method for generating router-level topologies. In our method, when a new node joins the network, the node likely connects to the nearest nodes. In addition, we add the new links based on the node utilization in the topology, which corresponds to an enhancement of network equipments in ISP networks. With appropriate parameters, important metrics, such as the a clustering coefficient and the amount of traffic that pass through nodes, exhibit the similar value of the actual ISP topology while keeping the degree distribution of resulting topology to follow power-law. We then apply the routing control method to the ISP topologies and show that the optimal routing method gives much smaller maximum link utilization (about 1/3) compared with the minimum hop routing which is often used in the operating networks. Accordingly, we examine a heuristic routing method suitable to the ISP topologies with consideration of technology constraints of IP routers. The evaluation results show that our modeling method can be actually used for evaluations on routing control.