Neubot: A Software Tool Performing Distributed Network Measurements to Increase Network Transparency

We present Neubot (the network neutrality bot), a network-measurement platform designed to run network-performance experiments from the network edges. The data produced by Neubot is useful to increase network transparency and to study network neutrality. We describe the Neubot architecture (based on plugins that emulate several protocols and are able to run client-server and peer-to-peer tests), which is one of the main contributions of this thesis. We describe the current Neubot implementation (Neubot 0.4.16.9), we provide up-to-date data concerning Neubot deployment, and we show how we used Neubot to run four diverse large-scale measurements campaigns involving more than 1,000 Neubot instances each. Such measurements campaign, which were only possible because the Neubot architecture was already flexible enough to allow us to deploy new network experiments on the already installed Neubot instances, were concerned with, respectively: the measurement of broadband speed using the HTTP protocol; the study of the link between application-level measurements and the packet-loss rate experienced by TCP (which is the other main contribution of this thesis); the study of rate adaptation algorithms for the dynamic adaptive streaming over HTTP streaming technology (DASH); emulating the BitTorrent protocol. We conclude the thesis with the description of Neuviz (the Neubot visualizer), a prototype data-visualization architecture that loads Neubot data and allows to navigate the data looking for potential deviations from network neutrality. Despite being still in beta stage, Neuviz already allowed to spot three anomalies in the median speeds measured by the Neubot `HTTP Speedtest' and BitTorrent tests.