SD-WAN Predictive Routing for Long-Term SLA Compliance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current predictive routing in SD-WANs operates on short-term scales, often failing to anticipate recurring conditions that cause traffic rerouting, leading to frequent SLA violations and poor user experience.
Innovation Solution
Implement predictive routing using risk and longevity metrics to make long-term predictions (LTPs) that consider historical data and evaluate the reliability and validity of network changes, filtering out risky or invalid predictions to optimize network configurations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning forecasting models are used for predictive routing, then traffic rerouting capability is improved, but the prediction validity period is limited to short-term scales
Solution Approach 1:
The system performs preliminary actions by making predictions about future network conditions before actual failures occur. The machine learning models analyze historical data and identify patterns that indicate potential SLA violations, allowing the system to proactively reroute traffic before the actual failure happens, thus extending the effective prediction horizon beyond simple short-term forecasting
Solution Approach 2:
The system continuously monitors network performance metrics and feeds this information back into the machine learning models. This feedback loop allows the models to learn from actual network behavior and adjust their predictions, enabling them to capture long-term patterns and improve prediction accuracy over extended periods rather than being limited to short-term forecasts
2Speed
If frequent traffic rerouting is performed based on short-term predictions, then response speed is improved, but SLA violations recur due to temporary fixes
Solution Approach 1:
The system performs preliminary analysis of network conditions using machine learning models to identify patterns that indicate upcoming SLA violations. By detecting these patterns in advance and making predictions about future network state, the system can proactively reroute traffic before failures occur, rather than reacting to short-term fluctuations, thus achieving both fast response and sustained reliability
Solution Approach 2:
The system changes the parameters used for routing decisions from short-term metrics to long-term predictive metrics. Instead of reacting to immediate network conditions, the system uses machine learning models to evaluate future network state based on historical patterns, transforming the routing decision-making process to focus on sustained SLA compliance rather than temporary fixes
3Productivity
If network changes are implemented based on predictions, then network optimization is improved, but risky or invalid predictions may cause deterioration
Solution Approach 1:
The system continuously monitors the impact of routing changes and feeds this information back into the machine learning models. This feedback mechanism allows the system to learn from the outcomes of previous predictions and adjustments, refining the models to distinguish between accurate and inaccurate predictions, thereby reducing the impact of risky or invalid predictions on network performance
Solution Approach 2:
The system introduces additional parameters for evaluating prediction quality, such as confidence intervals and validation metrics. By changing the decision-making process to incorporate these additional validation parameters, the system can filter out risky or invalid predictions before implementing network changes, thus improving both optimization efficiency and prediction reliability
Data Source
AI summary
In one embodiment, a device makes a prediction regarding service level agreement violations by a network transport available between a site and an online application. The device associates a risk metric with the prediction, based in part on a type of the network transport. The device computes a longevity metric for the prediction that indicates an expected validity period for the prediction. The device cause traffic to be routed between the site and the online application using the network transport, based on the prediction and its associated risk metric and its longevity metric.