Estimating the access link quality by active measurements

The access link quality experienced by the end users depends on the amount of traffic and on the presence of network anomalies. Different techniques exist to detect anomalies, but little attention has been devoted to quantify the access link quality and to which extent network anomalies affect the end user's access link experience. We refer to this aspect as the impact factor of the anomaly, that we define as the percentage of affected destinations. In the ideal case, a node should continuously monitor all possible routes to detect any degradation in performance, but this is not practical in reality. In this paper we show how a node can estimate the quality of Internet access through a limited set of measurements. We initially study the user's access network to understand the typical features of its connectivity tree. Then, we define an unbiased estimator for the quality of access and we compute the minimum number of paths to monitor, so that the estimator achieves a desirable accuracy without knowing the underlying topology. We use real data to construct a network graph and we validate our solution by causing a large number of anomalies and by comparing the real and the estimated quality of access for all available end hosts. Our results show that the impact factor is a meaningful metric to evaluate the quality of Internet access.