[GitHub] carsonwang commented on a change in pull request #20303: [SPARK-23128][SQL] A new approach to do adaptive execution in Spark SQL

[GitBox <gi...@apache.org>]

carsonwang commented on a change in pull request #20303: [SPARK-23128][SQL] A new approach to do adaptive execution in Spark SQL
URL: https://github.com/apache/spark/pull/20303#discussion_r246689468
 
 

 ##########
 File path: sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/QueryStage.scala
 ##########
 @@ -0,0 +1,222 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements.  See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License.  You may obtain a copy of the License at
+ *
+ *    http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.spark.sql.execution.adaptive
+
+import scala.concurrent.{ExecutionContext, Future}
+import scala.concurrent.duration.Duration
+
+import org.apache.spark.MapOutputStatistics
+import org.apache.spark.broadcast
+import org.apache.spark.rdd.RDD
+import org.apache.spark.sql.catalyst.InternalRow
+import org.apache.spark.sql.catalyst.expressions._
+import org.apache.spark.sql.catalyst.plans.physical.Partitioning
+import org.apache.spark.sql.execution._
+import org.apache.spark.sql.execution.exchange._
+import org.apache.spark.sql.execution.ui.SparkListenerSQLAdaptiveExecutionUpdate
+import org.apache.spark.util.ThreadUtils
+
+/**
+ * In adaptive execution mode, an execution plan is divided into multiple QueryStages. Each
+ * QueryStage is a sub-tree that runs in a single stage.
+ */
+abstract class QueryStage extends UnaryExecNode {
+
+  var child: SparkPlan
+
+  // Ignore this wrapper for canonicalizing.
+  override def doCanonicalize(): SparkPlan = child.canonicalized
+
+  override def output: Seq[Attribute] = child.output
+
+  override def outputPartitioning: Partitioning = child.outputPartitioning
+
+  override def outputOrdering: Seq[SortOrder] = child.outputOrdering
+
+  /**
+   * Execute childStages and wait until all stages are completed. Use a thread pool to avoid
+   * blocking on one child stage.
+   */
+  def executeChildStages(): Unit = {
+    // Handle broadcast stages
+    val broadcastQueryStages: Seq[BroadcastQueryStage] = child.collect {
+      case bqs: BroadcastQueryStageInput => bqs.childStage
+    }
+    val broadcastFutures = broadcastQueryStages.map { queryStage =>
+      Future { queryStage.prepareBroadcast() }(QueryStage.executionContext)
+    }
+
+    // Submit shuffle stages
+    val executionId = sqlContext.sparkContext.getLocalProperty(SQLExecution.EXECUTION_ID_KEY)
+    val shuffleQueryStages: Seq[ShuffleQueryStage] = child.collect {
+      case sqs: ShuffleQueryStageInput => sqs.childStage
+    }
+    val shuffleStageFutures = shuffleQueryStages.map { queryStage =>
+      Future {
+        SQLExecution.withExecutionId(sqlContext.sparkContext, executionId) {
+          queryStage.execute()
+        }
+      }(QueryStage.executionContext)
+    }
+
+    ThreadUtils.awaitResult(
+      Future.sequence(broadcastFutures)(implicitly, QueryStage.executionContext), Duration.Inf)
+    ThreadUtils.awaitResult(
+      Future.sequence(shuffleStageFutures)(implicitly, QueryStage.executionContext), Duration.Inf)
+  }
+
+  /**
+   * Before executing the plan in this query stage, we execute all child stages, optimize the plan
+   * in this stage and determine the reducer number based on the child stages' statistics. Finally
+   * we do a codegen for this query stage and update the UI with the new plan.
+   */
+  def prepareExecuteStage(): Unit = {
+    // 1. Execute childStages
+    executeChildStages()
+    // It is possible to optimize this stage's plan here based on the child stages' statistics.
+
+    // 2. Determine reducer number
+    val queryStageInputs: Seq[ShuffleQueryStageInput] = child.collect {
+      case input: ShuffleQueryStageInput => input
+    }
+    val childMapOutputStatistics = queryStageInputs.map(_.childStage.mapOutputStatistics)
+      .filter(_ != null).toArray
 
 Review comment:
   If the childStage's RDD has 0 partition, we will not submit that stage. See [ShuffleExchangeExec.eagerExecute](https://github.com/apache/spark/pull/20303/files#diff-3ceee31a3da1b7c71dddd32f666126fbR109). In that case, mapOutputStatistics will be null so we filter it.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org