Parallel Query Optimizers for Database Execution Plans

The patent divides the query optimization process into multiple independent optimizers, each specializing in different optimization techniques (cost-based optimization, rule-based optimization, heuristic optimization). Each optimizer processes queries independently and generates candidate execution plans, which are then evaluated to select the optimal plan. This segmentation allows each optimizer to focus on specific optimization strengths without being overwhelmed by the complexity of implementing all optimization strategies in a single system.

The patent implements a resource-aware optimizer selection mechanism that dynamically determines which optimizers to activate based on available system resources and query characteristics. When resources are constrained, not all optimizers are executed simultaneously; instead, a subset is selected based on expected performance benefits versus resource costs. This partial action approach ensures that multiple optimizers provide their optimization benefits while preventing excessive resource consumption that would occur if all optimizers ran regardless of system state.

The patent implements a hybrid approach that combines pre-computed statistical information (such as column cardinality statistics, data distribution histograms) with on-demand detailed analysis. Frequently accessed tables and columns maintain pre-analyzed statistics that can be quickly retrieved during query optimization. When higher precision is needed or statistics are outdated, the system performs targeted re-analysis rather than comprehensive re-computation of all statistics. This preliminary action approach ensures accurate optimization decisions while minimizing the time spent on data analysis.