Predictive Routing Models for SD-WAN QoS Reliability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional software-defined wide area networks (SD-WANs) rely on static routing metrics, leading to reactive failure detection and potential Quality of Service (QoS) issues due to periodic probing, which does not anticipate changes in network behavior, resulting in suboptimal traffic routing and reduced QoS.
Innovation Solution
Implementing predictive models that forecast network path behavior, using machine learning to compute routes based on probabilistic models as attributes, allowing for proactive rerouting before SLA violations occur, thereby enhancing QoS and reducing downtime.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If static routing metrics are used for route computation, then routing decisions are simple and fast, but the network cannot anticipate changes in path behavior leading to QoS degradation
Solution Approach 1:
The system performs preliminary actions by training machine learning models offline to predict path behaviors before actual traffic flows. The predictive models are pre-computed based on historical network data, allowing the routing system to make informed decisions without real-time computation overhead, thus improving QoS reliability while maintaining routing simplicity.
Solution Approach 2:
Machine learning predictive models serve as intermediaries between raw network data and routing decisions. These models translate complex network behavior patterns into predictable metrics that the routing system can use, bridging the gap between simple static metrics and complex dynamic behavior without requiring the routing computation to handle the full complexity.
2Reliability
If periodic probing is used to detect route changes, then failure detection is achieved, but traffic experiences reduced QoS in the meantime
Solution Approach 1:
The system performs preliminary detection by using predictive models to forecast path behavior changes before they actually occur. Instead of waiting for periodic probes to detect failures, the machine learning models anticipate QoS degradation in advance, allowing proactive rerouting that eliminates the QoS degradation period entirely.
Solution Approach 2:
The system takes preliminary anti-action by proactively rerouting traffic before SLA violations occur. The predictive models identify potential failures in advance, and the routing system counteracts the expected QoS degradation by switching paths beforehand, preventing the harmful effect rather than reacting to it after detection.
3Reliability
If reactive failure detection is implemented, then backup path rerouting is triggered, but the detection itself causes delay in maintaining QoS
Solution Approach 1:
The system performs preliminary detection of potential failures using machine learning predictive models that continuously assess path health based on historical patterns. This allows the system to identify at-risk paths before actual failures occur, enabling preemptive rerouting that eliminates the detection delay inherent in reactive approaches.
Data Source
AI summary
In one embodiment, a device obtains a predictive model that predicts a behavior of a path in a network. The device computes, based in part on the predictive model, a route in the network that includes the path, in accordance with a routing policy that instructs the device to use the predictive model as an attribute of the path during computation of the route. The device validates that the path exhibited the behavior predicted by the predictive model. The device initiates retraining of the predictive model, when the behavior predicted by the predictive model does not match the behavior of the path.