PoiEvent: An approach to extract the persistent and destructive routing events

To extract the persistent and destructive routing events is critical for Internet service providers (ISPs) to improve the network performance. Currently, the latest approach that leverages the notion of empathy to aggregate the paths that changed similarly over time can extract routing events from an arbitrary set of traceroutes. However, the cascading effects of path change prevent ISPs from accurately identifying the root cause of routing events. Meanwhile, the lack of evaluation for the impact of events limit ISPs from extracting the persistent and destructive routing events. In order to extract the persistent and destructive routing events and identify their root causes accurately, we propose PoiEvent. First, to infer the routing events and identify their root causes accurately, we improve the existing algorithm to remove the cascading effects. Then, to characterize the impact of routing events, we propose an event-based characterization method, which considers the location, severity, scope, congestion effect, and duration of each routing event. Finally, to extract the persistent and destructive routing events, we propose an event filtering method using the number of changed paths as hints to extract the routing events in terms of their impact. We perform experiments with data from RIPE Atlas to evaluate the performance of PoiEvent. The results show that PoiEvent can extract the persistent and destructive routing events. We believe that PoiEvent can be an effective aid for improving the network performance at the ISPs level.