Impact analysis of BGP sessions for prioritization of maintenance operations

When a limited number of network operators have to manage a large number of maintenance tasks, they often lose the sense of what is more important and end up focusing their time on lower-priority issues. As a result, operators may react slowly to critical tasks, increasing network downtime and maintenance costs. We propose a system that estimates the relative importance of different maintenance tasks. According to this estimation, network operators can prioritize their reaction, for example, by spending more time on avoiding disruption caused by high-impact tasks. In particular, we focus on configuration tasks related to BGP sessions. BGP sessions are frequently modified for maintenance operations, such as policy changes, router upgrades, and the addition of peers. These maintenance operations cause route changes, often leading to a huge amount of data loss. The proposed system estimates this amount of data loss by simulating the behavior of BGP. We implement the proposed system and estimate the impact of 372 sessions in a nationwide ISP network. We observe sessions with a wide range of impact, from those with nearly zero impact, to those that can result in 1000 GB of data loss if not properly protected. We also observe that these measures change over time, often in unpredictable ways (e.g., from 50 GB to 0 over a month period). According to this observation, we suggest that network operators perform periodic audits with the proposed system and classify sessions by highlighting those with a large impact. Operators can then prioritize the level of responses to the classified sessions accordingly and significantly reduce maintenance costs (i.e., by giving priority to the advanced protection of sessions with a large impact).