Efficient strategies for active interface-level network topology discovery
Baltra, Guillermo P.
MetadataShow full item record
As a piece of critical infrastructure, the Internet brings both benefits and security concerns. Recent cyber-security episodes such as route hijacks and Denial-of-Service attacks might have been mitigated and prevented with better knowledge of the network_s logical topology; i.e., router nodes and links. Current production public active mapping systems; e.g., Ark, Rocketfuel, and iPlane, produce valuable inferences of the Internet_s topology, as well as facilitating longitudinal analysis. We examine the extent to which the techniques utilized by these existing systems can be improved, in particular by attempting to reduce their high probing load. Our methodology divides the discovery process into three steps: destination selection, monitor assignment, and stop criterion. We implement and evaluate alternative designs for each step. The complete system runs in real-time on a production system to probe 500 randomly selected Internet subnetworks and gather real-world network maps. As compared to datasets from existing measurement platforms, we find that our method is able to generate 80% of the amount of data with 69% less load.