SD-WAN Predictive Routing via ML State Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional failure detection in SD-WANs is reactive, leading to network traffic disruption until rerouting occurs, and predicting what-if scenarios in SD-WANs is challenging due to complexity and system dynamics, making simulations difficult.
Innovation Solution
A system that constructs controlled what-if input parameters to predict network states using a network state model, which then feeds into a machine learning-based KPI prediction model to initiate routing changes based on predicted performance indicators.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If reactive failure detection using keep-alive mechanisms is used, then tunnel failures can be detected and traffic rerouted, but network traffic is disrupted until rerouting occurs
Solution Approach 1:
The system performs preliminary actions by proactively detecting potential tunnel failures using machine learning analysis of network metrics before the failure actually occurs. This allows the SD-WAN controller to pre-compute and prepare backup paths, so when a failure occurs, traffic can be rerouted immediately without disruption, thus resolving the contradiction between reliable failure detection and minimizing traffic disruption time.
2Reliability
If machine learning techniques are used for predictive failure detection, then proactive routing is enabled, but the complexity of training models to address what-if scenarios increases due to SD-WAN system dynamics
Solution Approach 1:
The system segments the complex SD-WAN network into multiple independent training domains or environments. Machine learning models are trained separately in each domain using domain-specific data and characteristics. This segmentation reduces the overall training complexity by breaking down the large, dynamic SD-WAN system into smaller, more manageable pieces that can be trained independently, while still enabling comprehensive predictive failure detection across the entire network.
Data Source
AI summary
In one embodiment, a device constructs a set of controlled what-if input parameters for evaluating a what-if scenario in a network. The device uses the set of controlled what-if input parameters and state data indicative of a current state of the network as input to a network state model. The network state model predicts values for the state data conditioned on the what-if input parameters. The device predicts a key performance indicator (KPI) in the network by using the predicted values for the state data from the network state model as input to a machine learning-based KPI prediction model. The device initiates a routing change in the network based in part on the predicted KPI.