From c2a01b8ca6581760d01d72ffe74770531918a015 Mon Sep 17 00:00:00 2001
From: Stephen Skeirik <skeirik2@illinois.edu>
Date: Sun, 1 Apr 2018 06:12:44 -0500
Subject: [PATCH] working on algorithm
---
.../scala/org/apache/spark/SparkEnv.scala | 2 +
.../CoarseGrainedExecutorBackend.scala | 6 ++
.../org/apache/spark/executor/Executor.scala | 8 +++
.../main/scala/org/apache/spark/rdd/RDD.scala | 64 +++++++++++++++++++
.../apache/spark/scheduler/DAGScheduler.scala | 61 +++++++++++++++++-
.../spark/storage/memory/MemoryStore.scala | 19 ++++++
updates | 30 +++++++++
7 files changed, 188 insertions(+), 2 deletions(-)
create mode 100755 updates
diff --git a/core/src/main/scala/org/apache/spark/SparkEnv.scala b/core/src/main/scala/org/apache/spark/SparkEnv.scala
index 72123f2232..1101063032 100644
--- a/core/src/main/scala/org/apache/spark/SparkEnv.scala
+++ b/core/src/main/scala/org/apache/spark/SparkEnv.scala
@@ -70,6 +70,8 @@ class SparkEnv (
val outputCommitCoordinator: OutputCommitCoordinator,
val conf: SparkConf) extends Logging {
+ var currentStage: Int = -1
+
private[spark] var isStopped = false
private val pythonWorkers = mutable.HashMap[(String, Map[String, String]), PythonWorkerFactory]()
diff --git a/core/src/main/scala/org/apache/spark/executor/CoarseGrainedExecutorBackend.scala b/core/src/main/scala/org/apache/spark/executor/CoarseGrainedExecutorBackend.scala
index 9b62e4b1b7..852666260e 100644
--- a/core/src/main/scala/org/apache/spark/executor/CoarseGrainedExecutorBackend.scala
+++ b/core/src/main/scala/org/apache/spark/executor/CoarseGrainedExecutorBackend.scala
@@ -95,6 +95,12 @@ private[spark] class CoarseGrainedExecutorBackend(
} else {
val taskDesc = TaskDescription.decode(data.value)
logInfo("Got assigned task " + taskDesc.taskId)
+ // SS {
+ val currentStageId = taskDesc.name.substring(taskDesc.name.lastIndexOf(' ') + 1, taskDesc.name.lastIndexOf('.')).toInt
+ // logEarne("Current in Stage: " + currentStageId)
+ env.currentStage = currentStageId
+ env.blockManager.currentStage = currentStageId
+ // SS }
executor.launchTask(this, taskDesc)
}
diff --git a/core/src/main/scala/org/apache/spark/executor/Executor.scala b/core/src/main/scala/org/apache/spark/executor/Executor.scala
index 2c3a8ef748..5738d15627 100644
--- a/core/src/main/scala/org/apache/spark/executor/Executor.scala
+++ b/core/src/main/scala/org/apache/spark/executor/Executor.scala
@@ -326,6 +326,14 @@ private[spark] class Executor(
throw new TaskKilledException(killReason.get)
}
+ // SS {
+ // logEarne("Task " + taskId + " is in Stage " + task.stageId + " and the depMap is " + task.depMap)
+ if (!env.blockManager.stageExInfos.contains(task.stageId)) {
+ env.blockManager.stageExInfos.put(task.stageId,
+ new StageExInfo(task.stageId, null, null, task.depMap, task.curRunningRddMap))
+ }
+ // SS }
+
// The purpose of updating the epoch here is to invalidate executor map output status cache
// in case FetchFailures have occurred. In local mode `env.mapOutputTracker` will be
// MapOutputTrackerMaster and its cache invalidation is not based on epoch numbers so
diff --git a/core/src/main/scala/org/apache/spark/rdd/RDD.scala b/core/src/main/scala/org/apache/spark/rdd/RDD.scala
index 0574abdca3..9f3e3273f9 100644
--- a/core/src/main/scala/org/apache/spark/rdd/RDD.scala
+++ b/core/src/main/scala/org/apache/spark/rdd/RDD.scala
@@ -214,6 +214,7 @@ abstract class RDD[T: ClassTag](
*/
def unpersist(blocking: Boolean = true): this.type = {
logInfo("Removing RDD " + id + " from persistence list")
+ sc.dagScheduler.renewDepMap(id)
sc.unpersistRDD(id, blocking)
storageLevel = StorageLevel.NONE
this
@@ -313,6 +314,34 @@ abstract class RDD[T: ClassTag](
ancestors.filterNot(_ == this).toSeq
}
+ // SS {
+ /**
+ * Return cached ancestor ids --- clone of above function.
+ */
+ private[spark] def getNarrowCachedAncestors: Set[Int] = {
+ val cachedAncestors = new mutable.HashSet[Int]
+ val ancestors = new mutable.HashSet[RDD[_]]
+ def visit(rdd: RDD[_]): Unit = {
+ val narrowDependencies = rdd.dependencies.filter(_.isInstanceOf[NarrowDependency[_]])
+ val narrowParents = narrowDependencies.map(_.rdd)
+ val narrowParentsNotVisited = narrowParents.filterNot(ancestors.contains)
+ narrowParentsNotVisited.foreach { parent =>
+ ancestors.add(parent)
+ if (parent.getStorageLevel != StorageLevel.NONE) {
+ cachedAncestors.add(parent.id)
+ } else {
+ visit(parent)
+ }
+ }
+ }
+
+ visit(this)
+
+ cachedAncestors.filterNot(_ == this.id).toSet
+ }
+ // SS }
+
+
/**
* Compute an RDD partition or read it from a checkpoint if the RDD is checkpointing.
*/
@@ -331,6 +360,41 @@ abstract class RDD[T: ClassTag](
private[spark] def getOrCompute(partition: Partition, context: TaskContext): Iterator[T] = {
val blockId = RDDBlockId(id, partition.index)
var readCachedBlock = true
+ // SS {
+ val blockManager = SparkEnv.get.blockManager
+
+ if (!blockManager.blockExInfo.containsKey(key)) {
+ blockManager.blockExInfo.put(key, new BlockExInfo(key))
+ }
+
+ blockManager.stageExInfos.get(blockManager.currentStage) match {
+ case Some(curStageExInfo) =>
+ var parExist = true
+ for (par <- curStageExInfo.depMap(rdd.id)) {
+ val parBlockId = new RDDBlockId(par, partition.index)
+ if (blockManager.blockExInfo.containsKey(parBlockId) &&
+ blockManager.blockExInfo.get(parBlockId).isExist
+ == 1) {
+ // par is exist
+
+ } else {
+ // par not exist now, add this key to it's par's watching set
+ parExist = false
+ if (!blockManager.blockExInfo.containsKey(parBlockId)) {
+ blockManager.blockExInfo.put(parBlockId, new BlockExInfo(parBlockId))
+ }
+ blockManager.blockExInfo.get(parBlockId).sonSet += key
+ }
+ }
+ if (parExist) {
+ // par are all exist so we update this rdd's start time
+ // logEarne("par all exist, store start time of " + key)
+ blockManager.blockExInfo.get(key).creatStartTime = System.currentTimeMillis()
+ }
+ case None =>
+ //logEarne("Some Thing Wrong")
+ }
+ // SS }
// This method is called on executors, so we need call SparkEnv.get instead of sc.env.
SparkEnv.get.blockManager.getOrElseUpdate(blockId, storageLevel, elementClassTag, () => {
readCachedBlock = false
diff --git a/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala b/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala
index 8c46a84323..77bed0cc62 100644
--- a/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala
+++ b/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala
@@ -162,6 +162,15 @@ class DAGScheduler(
private[scheduler] val activeJobs = new HashSet[ActiveJob]
+ // SS {
+ // Stores data on RDD dependencies for better caching algorithms
+ private[scheduler] var preRDDs = new HashSet[RDD[_]]
+
+ private[scheduler] var depMap = new HashMap[Int, Set[Int]]
+
+ private[scheduler] var curRunningRddMap = new HashMap[Int, Set[Int]]
+ // SS }
+
/**
* Contains the locations that each RDD's partitions are cached on. This map's keys are RDD ids
* and its values are arrays indexed by partition numbers. Each array value is the set of
@@ -944,6 +953,31 @@ class DAGScheduler(
logDebug("missing: " + missing)
if (missing.isEmpty) {
logInfo("Submitting " + stage + " (" + stage.rdd + "), which has no missing parents")
+ // SS {
+ // track dependency information when we submit a stage
+ env.blockManager.currentStage = stage.id
+ // logEarne("In " + stage + " we have cachedRDD " + sc.getPersistentRDDs)
+ // logEarne(stage + " has narrowAncestors " + stage.rdd.getNarrowAncestors)
+ // logEarne("Before " + stage + " we have RDDs as below: " + preRDDs)
+ val curRDDs = stage.rdd.getNarrowAncestors ++ Seq(stage.rdd)
+ val newRDDs = curRDDs.filter(!preRDDs.contains(_))
+ // logEarne("Stage " + stage + " will create RDD as below: " + newRDDs)
+ val newCachedRDDs = newRDDs.filter(_.getStorageLevel != StorageLevel.NONE)
+ // logEarne("Stage " + stage + " have newCachedRDDs as below: " + newCachedRDDs)
+ curRunningRddMap.clear()
+ newCachedRDDs.foreach { cachedRdd =>
+ depMap.put(cachedRdd.id, cachedRdd.getNarrowCachedAncestors)
+ curRunningRddMap.put(cachedRdd.id, cachedRdd.getNarrowCachedAncestors)
+ logEarne("this cachedRDD " + cachedRdd + " have below cachedAncestorRDDs: "
+ + cachedRdd.getNarrowCachedAncestors)
+ }
+ preRDDs = preRDDs ++ curRDDs
+ // logEarne("Current depMap is" + depMap)
+ if (!env.blockManager.stageExInfos.contains(stage.id)) {
+ env.blockManager.stageExInfos.put(stage.id,
+ new StageExInfo(stage.id, null, null, depMap.clone(), curRunningRddMap))
+ }
+ // SS }
submitMissingTasks(stage, jobId.get)
} else {
for (parent <- missing) {
@@ -957,6 +991,25 @@ class DAGScheduler(
}
}
+ // SS {
+ /** Renew the dependency map when an RDD is unpersisted */
+ def renewDepMap(id: Int): Unit = {
+ if (depMap.contains(id)) {
+ // logEarne("Remove RDD " + id + " from depMap")
+ val value = depMap(id)
+ depMap.foreach { rdd =>
+ if (rdd._2.contains(id)) {
+ val tmp = rdd._2 - id
+ depMap.put(rdd._1, tmp ++ value)
+ }
+ }
+ depMap.remove(id)
+ // logEarne("After Removed RDD " + id + " the depMap is " + depMap)
+ }
+ }
+ // SS }
+
+
/** Called when stage's parents are available and we can now do its task. */
private def submitMissingTasks(stage: Stage, jobId: Int) {
logDebug("submitMissingTasks(" + stage + ")")
@@ -1060,9 +1113,11 @@ class DAGScheduler(
val locs = taskIdToLocations(id)
val part = partitions(id)
stage.pendingPartitions += id
+ // SS {
new ShuffleMapTask(stage.id, stage.latestInfo.attemptNumber,
taskBinary, part, locs, properties, serializedTaskMetrics, Option(jobId),
- Option(sc.applicationId), sc.applicationAttemptId)
+ Option(sc.applicationId), sc.applicationAttemptId, depMap, curRunningRddMap)
+ // SS }
}
case stage: ResultStage =>
@@ -1070,9 +1125,11 @@ class DAGScheduler(
val p: Int = stage.partitions(id)
val part = partitions(p)
val locs = taskIdToLocations(id)
+ // SS {
new ResultTask(stage.id, stage.latestInfo.attemptNumber,
taskBinary, part, locs, id, properties, serializedTaskMetrics,
- Option(jobId), Option(sc.applicationId), sc.applicationAttemptId)
+ Option(jobId), Option(sc.applicationId), sc.applicationAttemptId, depMap, curRunningRddMap)
+ // SS }
}
}
} catch {
diff --git a/core/src/main/scala/org/apache/spark/storage/memory/MemoryStore.scala b/core/src/main/scala/org/apache/spark/storage/memory/MemoryStore.scala
index 4cc5bcb7f9..c5fdada55a 100644
--- a/core/src/main/scala/org/apache/spark/storage/memory/MemoryStore.scala
+++ b/core/src/main/scala/org/apache/spark/storage/memory/MemoryStore.scala
@@ -463,6 +463,25 @@ private[spark] class MemoryStore(
}
}
+
+ blockInfoManager.blockManager.inMemBlockExInfo.synchronized {
+ val bxi_iterator = blockManager.inMemBlockExInfo.iterator()
+ while (freedMemory < space && bxi_iterator.hasNext) {
+ val bxi = iterator.next()
+ if (blockIsEvictable(bxi.blockId,entries(
+ blockManager.stageExInfos.get(blockManager.currentStage) match {
+ case Some(curStageExInfo) =>
+ // cur is this stage's output RDD
+ if (!curStageExInfo.curRunningRddMap.contains(bxi.blockId.getRddId)) {
+ selectedBlocks += bxi.blockId
+ freedMemory += bxi.size
+ }
+ case None =>
+ // logEarne("ERROR HERE")
+ }
+ }
+ }
+
def dropBlock[T](blockId: BlockId, entry: MemoryEntry[T]): Unit = {
val data = entry match {
case DeserializedMemoryEntry(values, _, _) => Left(values)
diff --git a/updates b/updates
new file mode 100755
index 0000000000..9df2bcf24a
--- /dev/null
+++ b/updates
@@ -0,0 +1,30 @@
+Deleted
+====================================================
+spark/CacheManager :: relevant function (getOrCompute) moved into spark/rdd/RDD
+
+Moved:
+====================================================
+spark/Logging -> spark/internal/Logging :: added (logEarne) function which is just logging
+spark/storage/MemoryStore -> spark/storage/memory/MemoryStore :: modified tryToPut, ensureFreeSpace functions
+ 1. tryToPut was removed, subdivided between
+ MemoryManager.acquireStorageMemory and entries.synchronized - entries.put
+ 2. ensureFreeSpace was renamed to evictBlocksToFreeSpace
+
+Modified:
+====================================================
+spark/SparkEnv :: added currentStage variable
+spark/scheduler/DAGScheduler :: added preRDDs, depMap, curRunningRddMap --- based on narrowancestors and narrowcachedancestors, added renewDepMap function, added args to ShuffleMapTask/ResultTask
+spark/scheduler/ResultTask :: modified primary ctor to add argument
+spark/scheduler/ShuffleMapTask :: modified primary/aux ctor to add arguemnt
+spark/scheduler/Task :: modified primary/aux ctor to add arguemnt
+spark/rdd/RDD :: modified unpersist, added getNarrowCachedAncestors
+spark/executor/Executor :: for each new stage, store a new StageExInfo object in the environment in stageExInfos
+spark/executor/CoarseGrainedExecutorBackend :: set SparkEnv, BlockManager currentStage
+spark/storage/BlockId :: added getRddId, getRddSplitIndex functions
+spark/storage/BlockManager :: added blockExInfoMap, inMemBlockExInfo set, stageExInfos Map,
+ plus extra logic in removeRDD --- duh, this is because this is where the eviction happens!
+
+Created:
+====================================================
+spark/StageExInfo :: stores info related to a stage, i.e. alreadyPerRddSet, afterPerRddSet, depMap, curRunningRddMap
+spark/storage/BlockExInfo :: stores time to create block here, estimates de/serialization time
--
GitLab