Adaptive Query Execution (AQE)¶
Adaptive Query Execution (aka Adaptive Query Optimization, Adaptive Optimization, or AQE in short) is an optimization of a physical query execution plan in the middle of query execution for alternative execution plans at runtime.
Adaptive Query Execution re-optimizes the query plan based on runtime statistics.
Quoting the description of a talk by the authors of Adaptive Query Execution:
At runtime, the adaptive execution mode can change shuffle join to broadcast join if it finds the size of one table is less than the broadcast threshold. It can also handle skewed input data for join and change the partition number of the next stage to better fit the data scale. In general, adaptive execution decreases the effort involved in tuning SQL query parameters and improves the execution performance by choosing a better execution plan and parallelism at runtime.
InsertAdaptiveSparkPlan Physical Optimization¶
AQE Logical Optimizer¶
AQE Cost Evaluator¶
SparkPlan change happens,
AdaptiveSparkPlanExec prints out the following message to the logs:
Plan changed from [currentPhysicalPlan] to [newPhysicalPlan]
AQE QueryStage Physical Preparation Rules¶
Adaptive Query Execution notifies Spark listeners about a physical plan change using
Adaptive Query Execution uses logOnLevel to print out diagnostic messages to the log.
Adaptive Query Execution can change number of shuffle partitions and so is not supported for streaming queries (Spark Structured Streaming).