ML-Based Telemetry Classifier for SD-WAN Failure Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current failure detection in SD-WANs is reactive, leading to network traffic disruption until a failure is detected and rerouted, and collecting high-frequency telemetry data for machine learning-based predictive models is burdensome on network infrastructure.
Innovation Solution
A supervisory service that receives telemetry data samples from networking devices, trains a failure prediction model, assesses its performance, and sends a machine learning-based classifier to control which data samples are sent, optimizing data collection to reduce infrastructure load and improve predictive accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If high-frequency telemetry data is collected for machine learning-based predictive failure detection, then predictive accuracy is improved, but network infrastructure load and traffic overhead increase
Solution Approach 1:
The patent segments telemetry data collection by identifying and collecting only the most relevant parameters (such as CPU utilization, memory usage, disk I/O, network throughput) rather than all possible metrics. This selective collection reduces the volume of data transmitted to the machine learning system while maintaining predictive accuracy.
Solution Approach 2:
The patent extracts and transmits only the essential telemetry data samples needed for effective failure prediction, filtering out redundant or less informative data. The system extracts critical parameters from the full telemetry stream and sends only these to the machine learning service for analysis.
2Device complexity
If reactive failure detection is used, then implementation complexity is reduced, but traffic disruption duration increases
Solution Approach 1:
The patent implements preliminary action by using machine learning models to predict failures before they actually occur. The system analyzes telemetry data patterns to forecast potential failures and triggers proactive rerouting of traffic to backup paths, preventing traffic disruption rather than reacting after failure occurs.
Solution Approach 2:
The patent employs feedback mechanisms where the machine learning system continuously monitors telemetry data, compares it against learned patterns, and provides predictions about potential failures. This feedback loop enables the system to adjust traffic routing dynamically based on predicted network conditions.
Data Source
AI summary
In one embodiment, a supervisory service for one or more networks receives telemetry data samples from a plurality of networking devices in the one or more networks. The service trains a failure prediction model to predict failures in the one or more networks, using a training dataset comprising the received telemetry data samples. The service assesses performance of the failure prediction model. The service trains, based on the assessed performance of the failure prediction model, a machine learning-based classification model to determine whether a networking device should send a particular telemetry data sample to the service. The service sends the machine learning-based classifier to one or more of the plurality of networking devices, to control which telemetry data samples the one or more networking devices send to the supervisory service.