ML-Based Query Execution Time Estimator for RDBMS

The patent introduces a machine learning model as an intermediary component between the query execution plan and the resource allocation system. This ML model acts as a mediator that processes multiple complex factors (query complexity, system load, resource availability, lock conditions) and transforms them into accurate execution time predictions, resolving the contradiction by handling complexity internally while providing precise predictions to the user.

The patent replaces traditional mechanical/mathematical estimation methods with a machine learning-based predictive system. Instead of using deterministic formulas that struggle with multiple variables, the system uses trained ML models that can capture complex non-linear relationships between various factors and execution time, achieving higher prediction accuracy without increasing apparent system complexity.

The patent implements a feedback mechanism where execution time predictions from the ML model are continuously fed back to the resource allocation system. This feedback loop enables dynamic adjustment of resource allocation (CPU shares, I/O shares, memory) to ensure SLG achievement. The boosting module uses prediction results to automatically adjust resources, maintaining reliability while managing complexity through automated closed-loop control.

The patent performs preliminary resource allocation adjustments based on predicted execution times before queries actually execute. By using ML predictions to pre-determine required resource levels, the system can proactively allocate resources to meet SLG requirements, avoiding the need for complex real-time reactive adjustments during query execution.