Measurement of large-scale BGP events: Definition, detection, and analysis

We propose Large-scale BGP Event (LBE), which affects many IP prefixes (high impact) and is widely observable (non-local). To detect LBE, we propose the Update Visibility Matrix (UVM) to record the prefix and monitor related to each update. We formulate the problem of identifying LBE in UVM, which is NP-hard. Then we propose a heuristic algorithm to solve it. We apply the scheme to 2.18 TB of BGP updates and find that the identified LBEs are highly correlated with many well-known disruptive incidents. Besides, we identify 101 LBEs that have never been investigated before. By conducting case studies, we find that the LBEs have high impact and are caused by various reasons. Our work can assist in network/Internet management tasks such as problem prevention, diagnosis, and recovery.