Fixing BGP, one AS at a time

Debugging inter-domain routing problems on the Internet is notoriously hard. This is partly because BOP updates carry no information about the events that trigger them, and also because operation is highly distributed and complex, lacking a central point of control or authority. These factors have impeded the development of tools that can help in the diagnosis and troubleshooting of routing problems. Consequently, the dynamic behaviour of BGP is not well understood, even though it forms a critical component of the Internet infrastructure. In this paper, we argue that these problems can be effectively addressed by incorporating a diagnostic capability into the inter domain routing system. Such a capability would enable operators to effectively pinpoint the cause (and location) of routing problems and take timely steps to correct them. We present arguments to show that this capability is best realized by localized, AS specific entities called BGP Auditors. We then explore the high level design of a distributed, diagnostic framework and briefly discuss the relevant issues.