Package org.apache.flink.runtime.scheduler.adaptivebatch
package org.apache.flink.runtime.scheduler.adaptivebatch
-
ClassDescriptionThis scheduler decides the parallelism of JobVertex according to the data volume it consumes.Factory for
AdaptiveBatchScheduler.TheAdaptiveExecutionHandlerinterface defines the operations for handling the adaptive execution of batch jobs.A factory class for creating instances ofAdaptiveExecutionHandler.TheAdaptiveExecutionPlanSchedulingContextclass implements theExecutionPlanSchedulingContextinterface to provide a dynamic scheduling context that adapts execution plans based on runtime conditions.Information of All-To-All result.Internal configuration options for the batch job execution.Context for batch job recovery.Interface for handling batch job recovery.Utility class for bisection search.Helper class that provides read-only information of input forVertexParallelismAndInputInfosDecider.The blocking result info, which will be used to calculate the vertex parallelism and input infos.TheDefaultAdaptiveExecutionHandlerimplements theAdaptiveExecutionHandlerinterface to provide an incrementally generated job graph.Default implementation ofBatchJobRecoveryHandlerandJobEventReplayHandler.The default implementation ofSpeculativeExecutionHandler.Default implementation ofVertexParallelismAndInputInfosDecider.A dummy implementation of theBatchJobRecoveryHandler.The dummy implementation ofSpeculativeExecutionHandler.Interface for retrieving stream graph context details for adaptive batch jobs.This interface defines operations for components that are interested in being notified when new job vertices are added to the job graph.ANonAdaptiveExecutionHandlerimplements theAdaptiveExecutionHandlerinterface to provide an immutable static job graph.TheNonAdaptiveExecutionPlanSchedulingContextis a final class that implements a scheduling context for execution plans that do not require adaptive changes.This class represents the information about the finished operators.Information of Pointwise result.Interface for managing speculative execution of tasks and handling slow task detection.Defines an optimization strategy for StreamGraph.TheStreamGraphOptimizerclass is responsible for optimizing a StreamGraph based on runtime information.VertexParallelismAndInputInfosDecideris responsible for deciding the parallelism andJobVertexInputInfos of a job vertex, based on the information of the consumed blocking results.