Measuring the Impacts of Sampling Bias on Internet AS-Level Topology Inference

In this paper, we carried out a systematic qualitative study of the impacts of sampling bias over Internet AS-level topology inference. Based on the relatively complete Chinese Internet AS graph derived from our large-scale measurement effort, we further quantified such impacts by showing how much peer-peer and provider-customer links were missed due to sampling bias. Moreover, we brought to light the possible incorrect inferences that can be derived about relationships between ASes if the number of sampling points is limited.