Contextual Multi-Armed Bandits for User-Centric Network Path Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting and managing network paths in SD-WANs fail to accurately reflect user experience due to reliance on synthetic probing, which often misses micro failures and dynamic user experiences, leading to SLA violations and disruptions.
Innovation Solution
Implementing a multi-armed bandit model to learn from real user experience metrics and select optimal network paths using contextual multi-armed bandits to predict and avoid disruptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If path probing is used to predict network performance, then predictive failure detection becomes possible, but the probing results do not accurately reflect true user experience with the application
Solution Approach 1:
The patent introduces an intermediary layer between path probing and routing decisions. Instead of directly using probing results, the system uses machine learning models to translate synthetic probing data into predicted user experience metrics. This intermediary translation layer allows the system to maintain the benefits of proactive probing while improving the accuracy of user experience prediction by learning the complex mapping between synthetic and real user experience.
Solution Approach 2:
The system creates a virtual copy of the user experience through machine learning models. Rather than directly measuring real user experience (which is difficult to obtain), the system probes the network and uses ML to generate a predictive copy of what actual users will experience. This copying approach enables proactive failure detection while maintaining reasonable accuracy in predicting true user experience.
2Loss of information
If all available paths are probed to determine optimal routing, then comprehensive path information is obtained, but micro failures and dynamic user experiences are not captured
Solution Approach 1:
The system performs preliminary probing actions on all available paths to gather baseline performance data before actual user traffic arrives. By proactively probing paths in advance and using machine learning to predict future performance, the system can identify potential micro failures and performance degradation before they impact real users, enabling preventive routing decisions.
Solution Approach 2:
The system implements continuous feedback loops where probing results, actual user experience data, and performance metrics are fed back into the machine learning models. This feedback mechanism allows the system to learn from both comprehensive path probing data and real user experience, improving its ability to detect micro failures and predict future performance across all paths.
3Extent of automation
If synthetic probing is used for network path selection, then automated routing decisions can be made, but SLA violations and user experience disruptions occur
Solution Approach 1:
The patent replaces the mechanical/probing-based routing decision system with an intelligent machine learning-based system. Instead of relying solely on synthetic probing metrics, the system uses ML models to predict actual user experience and SLA compliance. This substitution transforms automated routing from a purely synthetic-metric-driven process to one that predicts real user experience, thereby improving SLA compliance while maintaining automation.
Data Source
AI summary
In one embodiment, a device uses a multi-armed bandit model to select different network paths over time via which traffic associated with an online application is routed. The device obtains, from a provider of the online application, application experience metrics associated with the different network paths and indicative of user satisfaction with the online application. The device learns, by the multi-armed bandit model, which of the different network paths will provide satisfactory application experience metrics, based on the application experience metrics associated with the different network paths. The device causes the traffic associated with the online application to be routed via a set of one or more paths expected by the multi-armed bandit model to provide satisfactory application experience metrics for the online application.