Determining the Cause and Frequency of Routing Instability with Anycast

摘要

In this article we present a methodology by which an autonomous system (AS) can estimate the stability of their BGP routes without requiring access to restricted BGP data. We demonstrate a novel measurement approach using DNS anycast as an indicator of routing instability. Using this method, even end-users may monitor their ISP's routing stability, something which was previously infeasible without the continual use of expensive ICMP traceroutes or access to generally restricted routing information. We then perform a case study from within a large ISP in order to quantify and determine the cause of the routing instability. To determine causation, we correlate external and internal BGP events with variations in the final anycast destination. We conclude that anycast is extremely sensitive to anomalous BGP events and that by monitoring anycast it may be possible for large networks to receive early warning of BGP instability.