Nearline Content Playback Optimization for Transient Network Conditions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing content delivery systems face challenges in predicting and adjusting to transient and unstable network conditions, leading to suboptimal quality of experience (QoE) due to the limitations of historical data in capturing recent network variations.
Innovation Solution
A prediction model that incorporates nearline information, dynamically trained with recent data, to adjust content delivery processes such as network selection and adaptive bitrate algorithms, improving QoE by reacting to current network conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If historical data is used for training prediction models, then the model has sufficient training data, but it cannot capture recent network variations and transient conditions
Solution Approach 1:
The patent combines historical data with nearline data to train prediction models. Historical data provides sufficient training samples while nearline data captures recent network variations. The merging of these two data sources allows the system to maintain reliable predictions while adapting to current network conditions, resolving the contradiction between having enough training data and capturing recent variations.
Solution Approach 2:
The system dynamically adjusts the prediction model by incorporating nearline data that reflects current network conditions. Instead of using static historical data alone, the model becomes dynamic by integrating nearline information about recent network variations, allowing it to adapt to changing conditions while maintaining prediction reliability.
2Adaptability or versatility
If the prediction model is updated frequently with nearline data, then it reacts to current network conditions, but it increases system complexity and computational overhead
Solution Approach 1:
The system applies partial updating by selectively incorporating nearline data rather than completely retraining the model frequently. It uses a portion of nearline data to adjust predictions without performing full model retraining, thus achieving adaptability to current network conditions while avoiding the excessive computational overhead of frequent complete updates.
Solution Approach 2:
The system performs preliminary actions by pre-processing and storing nearline data in a structured format before it is needed for prediction updates. This preliminary organization of nearline information reduces the computational burden during actual prediction updates, allowing frequent adaptations without proportionally increasing system complexity.
3Measurement precision
If historical features are aggregated for quality of experience prediction, then comprehensive analysis is achieved, but transient network issues are not detected timely
Solution Approach 1:
The system implements feedback mechanisms by continuously monitoring nearline network conditions and using this information to adjust quality of experience predictions in real-time. The feedback loop compares historical predictions with actual nearline performance, enabling timely detection of transient network issues while maintaining comprehensive analysis through aggregated historical features.
Solution Approach 2:
The system maintains continuous monitoring and updating of predictions by seamlessly integrating historical feature aggregation with nearline data processing. This continuous action ensures both comprehensive historical analysis and timely detection of transient issues, as the system never stops collecting or processing data but transitions smoothly between historical and nearline information.
Data Source
AI summary
In some embodiments, a method determines training data from first nearline features for a first sliding time window and a fixed time window. The first sliding time window changes during a first time period and the fixed time window is static during the first time period. A prediction model is trained using the training data. The prediction model is trained for a second time period in the first time period. The method determines values for second nearline features for a second sliding time window and a fixed time window for a request for a current session that is received during the second time period. The second sliding time window is based on a time for the request. The values are input for the second nearline features into the prediction model to generate a prediction. The method performs an action for the current session based on the prediction.