Dynamic BFD Echo Session Distribution Across Multi-Processor Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current network protocols face delays in detecting link failures, which can lead to significant downtime and performance reduction in critical applications, as they rely on slower 'hello' mechanisms or higher-level protocols that take at least one second to detect failures, whereas lower-level protocols like BFD can detect failures in under 5 milliseconds but may generate false positives during reassignment of BFD sessions.
Innovation Solution
Implementing a Bidirectional Forwarding Detection (BFD) module that dynamically distributes echo sessions across multiple nodes based on criteria such as node behavior and availability, using dormant BFD clients to minimize false positives and reduce recovery time by quickly reassigning server roles, thereby reducing the window for false positive detections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If BFD echo sessions are concentrated on a single node for fast failure detection, then detection speed is improved, but false positive rate increases during node reassignment
Solution Approach 1:
The patent divides BFD echo sessions across multiple nodes (segmentation of detection function) so that no single node holds all sessions. This distributes the detection load and ensures that during node reassignment, not all sessions are affected simultaneously, reducing false positives while maintaining fast detection through the distributed architecture
Solution Approach 2:
The patent pre-assigns BFD echo sessions to multiple potential server nodes in advance (preliminary distribution). When a node fails or needs reassignment, the sessions are already positioned to be quickly transferred to alternative nodes, minimizing the reassignment window and reducing false positives without sacrificing detection speed
2Loss of time
If BFD sessions are quickly reassigned during node failure, then recovery time is reduced, but the window for false positive detection increases
Solution Approach 1:
The system performs preliminary distribution of BFD sessions across multiple nodes before any failure occurs. This pre-positioning allows for immediate failover without extended reassignment windows, as the alternative nodes are already prepared and configured to handle the sessions, thus reducing both recovery time and false positive opportunities
Solution Approach 2:
The patent implements dynamic reassignment of BFD sessions among multiple nodes based on real-time node availability and status. This dynamic approach allows the system to adaptively transfer sessions during failures while maintaining optimal distribution, reducing the time nodes remain in transitional states where false positives could occur
Data Source
AI summary
A method and system of failure detection in a computer network, including pre-programming a number of processors of a multi-processor networked system coupled with remote systems via a network, each of the processors configured to run a BFD module. One processor may be assigned to be an active BFD server to establish sessions between the multi-processor networked system and the remote systems to detect link faults between the multi-processor networked system and the remote systems. Other processors may be assigned as dormant BFD clients. If a link or a processor running the active BFD server fails, one of the dormant BFD clients is selected to become an active BFD server. The BFD server may be dynamically replaced with a dormant client based on predetermined criteria without failure of the BFD server.