Periodic Path Changes in RIPE Atlas

Abstract

Large-scale data sets of the Internet measurements are commonly used by researchers and operators for investigating Internet performance or for tackling network issues. Looking at sequences of traceroutes in such data sets, it is common to observe paths that change over time. We are interested in verifying if there are periodic phenomena affecting such path changes and, if yes, in determining if they depend on artifacts of the used data set or on topology changes of the network. For this purpose, we devise a novel algorithm for detecting periodicities in sequences of traceroutes. Then, we exploit the algorithm for analyzing the traceroutes produced by the RIPE Atlas, a popular public measurement platform. We study and report the features of the found periodicities and some of their causes. We found that: 1) a surprisingly large percentage of the traceroutes exhibit a periodic behavior; 2) a large number of periodicities depend on the RIPE Atlas platform itself; and 3) a smaller amount is related to the MPLS and load balancing.