Efficient IP-Level network topology capture

Abstract

Large-scale distributed network route tracing systems obtain the IP-level internet topology and can be used to monitor and understand network behavior. However, existing approaches require one or more days to obtain a full graph of the public IPv4 internet, which is too slow to capture important network dynamics. This paper presents a new approach to topology capture that aims at obtaining the graph rather than full routes, and that employs partial rather than full route tracing to achieve this aim. Our NTC (Network Topology Capture) heuristics use information from previous tracing rounds to guide probing in future rounds. Through simulations based upon two months of traces that we obtained, we find that the heuristics improve significantly on the state of the art for reducing probing overhead while maintaining good graph coverage. We also conduct the first study of how such a distributed tracing system performs in its ability to capture network dynamics. © 2013 Springer-Verlag Berlin Heidelberg.