SD-WAN HA Switchover Using Exit State Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing SDWAN environments face challenges in accurately predicting failure severity due to the use of individual failure types for switchover decisions, leading to inconsistent and nondeterministic outcomes that can result in black holing or application degradation.
Innovation Solution
Implement a method where HA routers communicate their exit statuses to each other and apply an algorithm based on local and far-end exit statuses, using a state table to determine switchover decisions, rather than relying on individual failure decrements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If individual failure types are used for switchover decisions, then the system can respond to specific failures, but the switchover decisions become inconsistent and nondeterministic
Solution Approach 1:
The patent segments the failure assessment into two independent components: local exit status (assessing the active router's own exits) and far-end exit status (assessing the standby router's exits). Each component is evaluated separately using standardized criteria, and their combination through an algorithm produces deterministic switchover decisions. This segmentation eliminates the inconsistency caused by individual failure type decrements.
Solution Approach 2:
The patent changes the assessment parameters from individual failure type decrements to a comprehensive evaluation of exit statuses (up/down states) of DIA and WAN exits. By transitioning to binary status parameters (exit up/exit down) rather than cumulative decrement values, the system achieves both precision in failure detection and consistency in switchover decisions.
2Device complexity
If stateful features run only on active node, then the system simplifies state management, but the standby node cannot take over effectively on failure
Solution Approach 1:
The patent implements preliminary action by having the standby node continuously assess and monitor the far-end exit statuses in advance. This allows the standby node to be pre-prepared with knowledge of the active node's exit conditions, enabling it to take over effectively when needed without requiring complex state synchronization at failover time.
Solution Approach 2:
The patent establishes a feedback mechanism where both active and standby nodes continuously exchange and assess each other's exit statuses. This bidirectional feedback ensures that both nodes have current information about the system state, allowing the standby to effectively take over while maintaining simplified state management through standardized status reporting.
3Reliability
If deterministic switchover decisions are implemented, then the reliability improves, but the system requires comprehensive communication between HA routers
Solution Approach 1:
The patent applies universality by designing a communication mechanism that serves multiple functions: exit status reporting, failover coordination, and system state synchronization. The same communication channels used for standard HA protocols are leveraged to transmit the comprehensive exit status information needed for deterministic decisions, avoiding the need for separate dedicated communication infrastructure.
Data Source
AI summary
In one aspect, a method includes determining, at a first node in an active state, a local wide area network (WAN) tunnel state of a WAN tunnel of the first node based on a forwarding table of the first node; determining, at the first node, a local direct internet access (DIA) path state of a DIA path of the first node based on a DIA tracker; receiving, at the first node, state information associated with a second node, wherein the state information includes a far-end WAN tunnel state and a far-end DIA path state associated with the second node; comparing, at the first node, the local WAN tunnel state and the local DIA path state with the far-end WAN tunnel state and the far-end DIA path state of the second node, respectively to yield a comparison; and based on the comparison, switching, by the first node, to a standby state.