The Internet Pendulum: On the periodicity of Internet Topology Measurements

Abstract

Public databases of large-scale topology measures (e.g. RIPE Atlas) are very popular both in the research and in the practitioners communities. They are used, at least, for understanding the state of the Internet in real time, for outage detection, and to get a broad baseline view of the Internet evolution over time. However, despite the large amount of investigations, the dynamic aspects of these measures have not been fully understood. As an example, looking at time-series of such measures it happens to observe patterns that repeat at regular intervals. More specifically, looking at a time-series of traceroutes involving certain source-target pairs it happens to observe that the paths follow alternations that repeat several times. Have they the features of periodicity? What are their main characteristics? In this paper we study the problem of detecting and characterizing periodicities in Internet topology measures. For this purpose we devise an algorithm based on autocorrelation and string matching. First, we validate the effectiveness of our algorithm in-vitro, on randomly generated measures containing artificial periodicities. Second, we exploit the algorithm to verify how frequently traceroute sequences extracted from popular databases of topology measures exhibit a periodic behavior. We show that a surprisingly high percentage of measures present one or more periodicities. This happens both with traceroutes performed at different frequencies and with different types of traceroutes. Third, we apply our algorithm to databases of BGP updates, a context where periodicities are even more unexpected than the one of traceroutes. Also in this case our algorithm is able to spot periodicities. We argue that some of them are related to oscillations of the BGP control plane.