Predictive Routing Policy Splitting in SD-WAN
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional software-defined wide area networks (SD-WANs) rely on reactive failure detection and routing, which is inefficient as it only reroutes traffic after a failure is detected, whereas predictive failure detection and proactive routing are not effectively implemented due to the complexity of automatically generating and managing routing policy changes.
Innovation Solution
A device splits a software-defined network's routing policy into multiple policies based on predictive routing forecasts, evaluates whether to revert these policies, and sends data to a user interface for evaluation and potential permanent changes, using machine learning models to anticipate and prevent service level agreement (SLA) violations by proactively rerouting traffic.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If reactive failure detection and routing is used, then the system is simple to implement, but the reliability and service continuity deteriorate because traffic is only rerouted after failure detection
Solution Approach 1:
The system performs preliminary actions by using machine learning models to predict potential tunnel failures before they occur. The predictive model analyzes historical performance data, traffic patterns, and network conditions to forecast failures, allowing the SD-WAN controller to proactively reroute traffic through alternative tunnels before the predicted failure occurs, thus maintaining service continuity without waiting for actual failure detection
Solution Approach 2:
The patent introduces an intermediary machine learning-based predictive model that acts as a mediator between network monitoring data and routing decisions. This predictive layer processes raw network metrics and generates failure probability forecasts, which then inform the routing policy adjustments. The intermediary model translates complex network states into actionable predictions, enabling intelligent proactive routing without requiring direct complex rule-based systems
2Productivity
If automatic routing policy changes are implemented based on predictive failures, then the productivity and response time improve, but the device complexity and difficulty of managing routing policies increase
Solution Approach 1:
The system implements self-service by enabling automatic generation and application of routing policy changes based on predictive failure signals. When the machine learning model predicts a tunnel failure, the SD-WAN controller automatically generates appropriate routing policies to divert traffic through alternative paths without requiring manual intervention. The system monitors the outcomes and can automatically adjust or revert policies based on actual performance, making the routing management self-regulating
Solution Approach 2:
The patent applies parameter changes by dynamically adjusting routing policy parameters based on predictive failure probabilities. Instead of static routing rules, the system modifies routing parameters such as path selection, traffic distribution weights, and tunnel priorities in real-time based on ML model outputs. These parameter adjustments enable flexible adaptive routing that responds to predicted network conditions while maintaining manageable policy structures through systematic parameter control
3Adaptability or versatility
If routing policies are split into multiple policies based on forecasts, then the adaptability and precision of routing decisions improve, but the ease of operation and user complexity increase
Solution Approach 1:
The patent implements universality by creating a unified predictive routing framework that handles multiple routing scenarios through a single integrated system. The machine learning model generates comprehensive forecasts that inform various routing decisions across different tunnel types, traffic classes, and failure scenarios. The SD-WAN controller applies universal routing policies that can adapt to multiple situations, reducing the need for separate specialized policies for each scenario while maintaining high adaptability
Data Source
AI summary
In one embodiment, a device obtains routing forecasts for a software defined network. The device splits a particular routing policy for the software defined network into two or more routing policies, based on the routing forecasts. The device makes an evaluation as to whether the two or more routing policies should be reverted back into the particular routing policy. The device sends, to a user interface, data indicative of the particular routing policy that was split into the two or more routing policies and the evaluation as to whether the two or more routing policies should be reverted back into the particular routing policy.