Dynamic Boost Time Management for Application Launch

The patent applies dynamics by transitioning from fixed boosting duration to dynamic boosting duration that adapts based on application characteristics and system state. The AI model predicts the optimal boosting duration for each specific application launch scenario, allowing the system to adjust the boost time dynamically rather than using a predetermined fixed value, thereby avoiding both over-boosting and under-boosting.

The patent changes the parameter of boosting duration from a static fixed value to a dynamic predicted value based on multiple factors including application type, system load, and hardware capabilities. The AI model computes the optimal boosting duration by analyzing various parameters and their interactions, enabling precise control over the boost period to match actual application requirements.

The patent implements self-service by enabling the system to automatically determine the optimal boosting duration for each application launch without requiring manual configuration or intervention. The AI model autonomously analyzes application characteristics, system state, and historical data to compute the precise boost duration needed, making the system self-adjusting and eliminating the need for external tuning while maintaining high performance accuracy.

The patent incorporates feedback mechanisms where the system continuously monitors application launch performance and uses this information to refine future boosting decisions. The AI model learns from past launch scenarios and adjusts its predictions accordingly, creating a closed-loop system that improves performance accuracy over time while maintaining operational simplicity.

The patent applies partial action by selectively invoking the AI model only when necessary rather than continuously. The system assesses whether AI prediction is needed based on current system conditions, application characteristics, and confidence levels from simpler prediction methods. This approach achieves accurate predictions when required while avoiding the resource overhead of continuous AI model execution, thereby preserving system health.

The patent implements local quality by applying different prediction strategies to different application scenarios. For some applications or system states, simpler and less resource-intensive prediction methods are used, while for others requiring higher precision, the full AI model is invoked. This localized approach ensures prediction accuracy is optimized only where necessary, reducing overall system resource consumption and maintaining system health.