BGP Peer Dampening Mechanism for Flapping Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
BGP connection restarts, known as 'flapping', consume significant network bandwidth and CPU resources due to the need for full Adjacent-RIB exchanges, often resulting from configuration errors or hardware issues, and can destabilize the network.
Innovation Solution
Implementing a 'dampening' mechanism that tracks the history of flapping events, penalizes unstable peers, and prevents automatic reconnection when flapping events exceed a defined threshold, allowing the network to stabilize and conserve resources by bypassing unstable links.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If BGP connection restarts are allowed to maintain routing updates, then routing information can be refreshed, but network bandwidth and CPU resources are consumed significantly
Solution Approach 1:
The patent applies parameter changes by introducing a penalty parameter that accumulates with each flapping event. When the penalty exceeds a threshold, the BGP peer is dampened and connection restarts are prevented. This transforms the binary decision of allowing or blocking restarts into a dynamic parameter-based control system that adapts to the stability history of the peer relationship.
Solution Approach 2:
The dampening mechanism implements feedback by monitoring the stability history of BGP peer connections and using this information to control future connection behavior. The penalty accumulator provides continuous feedback about the peer's stability, and this feedback loop allows the system to automatically adjust its behavior to prevent resource-consuming flapping while maintaining necessary routing updates.
2Stability of the object's composition
If BGP session restarts are prevented to conserve resources, then network stability is improved, but routing information may become outdated
Solution Approach 1:
The dampening mechanism is dynamic rather than static. It allows connection restarts when the penalty is below the threshold (indicating stable peers) and prevents restarts when the penalty exceeds the threshold (indicating unstable peers). This dynamic behavior ensures that routing information is maintained for stable peers while preventing resource exhaustion from unstable peers, achieving a balance between stability and information freshness.
Solution Approach 2:
The patent segments the BGP peer population into different categories based on their stability characteristics. Peers with low penalty values are treated differently from those with high penalty values. This segmentation allows the system to apply different policies to different peers, maintaining routing information for stable peers while conserving resources for unstable ones.
3Reliability
If full Adjacent-RIB exchanges are performed on each restart, then complete routing information is synchronized, but network bandwidth is consumed excessively
Solution Approach 1:
Instead of performing full Adjacent-RIB exchanges on every restart attempt, the dampening mechanism applies partial action by preventing restarts for peers that have exceeded the penalty threshold. This selective approach performs routing synchronization only when necessary (for stable peers) and skips it for unstable peers, significantly reducing overall bandwidth consumption while maintaining routing information completeness for reliable peers.
Data Source
AI summary
A method is disclosed for preventing an unstable BGP Peer from repeatedly initializing unstable BGP connections. In one embodiment, BGP speakers are penalized for causing errors that result in BGP restarts. When a speaker accumulates enough penalty points, its peer notifies it that it has been dampened (prevented from establishing a BGP connection). A memory decay function allows the speaker to automatically attempt a new connection once a given amount of time has passed. The method allows at least two, and possibly more, BGP speakers to avoid network and processor costs from servicing unstable BGP peerings.