Reinforcement Learning Probing for QoE Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for assessing quality of experience (QoE) in computer networks rely on proxies like Service Level Agreements (SLAs) which are often unreliable, as they do not accurately reflect user experience, especially in cases of sharp spikes in network metrics that average out over time.
Innovation Solution
Implementing a device that uses reinforcement learning to adjust probing strategies based on the predictive model's performance in measuring QoE, optimizing the strategy to find a minimally disruptive approach that provides accurate performance by performing probing tests in the network and adjusting the strategy accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If SLA violations are used as a proxy for QoE assessment, then the assessment process is simple and straightforward, but the accuracy of QoE prediction deteriorates because SLA violations do not accurately reflect true user experience
Solution Approach 1:
The patent replaces the traditional mechanical/procedural approach of using SLA violation checks with a machine learning-based predictive model. The system collects network telemetry data (packet loss, latency, jitter) and uses trained machine learning models to predict actual QoE scores, substituting the simple but inaccurate SLA proxy method with a more sophisticated data-driven approach that accurately reflects user experience.
2Measurement precision
If comprehensive telemetry is collected to improve QoE prediction accuracy, then the accuracy of the predictive model improves, but the network overhead increases
Solution Approach 1:
The patent implements adaptive telemetry collection where the system gathers more detailed network metrics when QoE degradation is detected or suspected, and reduces telemetry collection when network conditions are stable and QoE is satisfactory. This partial action approach ensures sufficient data is collected to maintain prediction accuracy while avoiding continuous exhaustive data collection that would create excessive network overhead.
Solution Approach 2:
The system dynamically adjusts telemetry collection parameters such as sampling rate, data granularities, and metric types based on current network conditions and QoE prediction needs. When network conditions are stable, the system reduces the frequency and detail of telemetry collection; when degradation is detected, it increases collection intensity, thereby optimizing the balance between model accuracy and network overhead.
Data Source
AI summary
In one embodiment, a device causes, in accordance with a probing strategy, performance of a probing test by one or more agents in a network and with respect to an online application. The device obtains quality of experience measurements for the online application. The device adjusts, using reinforcement learning, the probing strategy based on how well a predictive model was able to predict the quality of experience measurements given results of the probing test. The device repeats the causing, obtaining, and adjusting steps using the probing strategy adjusted by the device, to find a minimally disruptive probing strategy that provides acceptable performance by the predictive model.