Abstract :
[en] Since the early 2000's, the research community has explored many approaches to discover and study the Internet topology, designing both data collection mechanisms and models.
In this paper, we introduce SAGE (Subnet AggrEgation), a new topology discovery tool that infers the hop-level graph of a target network from a single vantage point. SAGE relies on subnet-level data to build a directed acyclic graph of a network modeling how its (meshes of) routers, a.k.a. neighborhoods, are linked together. Using two groundtruth networks and measurements in the wild, we show SAGE accurately discovers links and is consistent with itself upon a change of vantage point.
By mapping subnets to the discovered links, the directed acyclic graphs discovered by SAGE can be re-interpreted as bipartite graphs. Using data collected in the wild from both the PlanetLab testbed and the EdgeNet cluster, we demonstrate that such a model is a credible tool for studying computer networks.
