MAXIMUM LIKELIHOOD IDENTIFICATION OF NETWORK TOPOLOGY FROM END-TO-END MEASUREMENTS

One of the predominant schools of thought in networking toda y is that monitoring and control of large scale networks is only p ractical at the edge. With intelligent and adaptive elements at the ed g of the network, core devices can function as simple, robust router s. However, the effectiveness of edge-based control can be signifi ca tly enhanced by information about the internal network state. If t he core is endowed with minimal monitoring and data collection capa bilities, then methods for inferring state information from edg -based traffic measurements are of great interest. One of the most fu ndamental components of the state is the routing topology. The f ocus of this paper is a new Maximum Likelihood approach to topolog y identification that makes use only of measurements performe d between host computers and requires no special support (e.g., ICMP responses) from internal routers.