概要
Spark 1.1中对spark core的一个重大改进就是引入了sort-based shuffle处理机制，本文就该处理机制的实现进行初步的分析。
Sort-based Shuffle之初体验
通过一个小的实验来直观的感受一下sort-based shuffle算法会产生哪些中间文件，具体实验步骤如下所述。
步骤1： 修改conf/spark-default.conf, 加入如下内容
spark.shuffle.manager SORT
步骤2: 运行spark-shell
SPARK_LOCAL_IP=127.0.0.1 $SPARK_HOME/bin/spark-shell
步骤3: 执行wordcount
sc.textFile("README.md").flatMap(l => l.split(" ")).map(w=>(w,1)).reduceByKey(_ + _).collect
步骤4: 查看生成的中间文件
find /tmp/spark-local* -type f
文件查找结果如下所示
/tmp/spark-local-20140919091822-aa66/0f/shuffle_0_1_0.index
/tmp/spark-local-20140919091822-aa66/30/shuffle_0_0_0.index
/tmp/spark-local-20140919091822-aa66/0c/shuffle_0_0_0.data
/tmp/spark-local-20140919091822-aa66/15/shuffle_0_1_0.data
可以看到生成了两人种后缀的文件，分别为data和index类型，这两者的用途在后续分析中会详细讲述。
如果我们做一下对比实验，将shuffle模式改为Hash，再来观看生成的文件，就会找到区别。将原先配置文件中的SORT改为HASH，重新启动spark-shell，执行相同的wordcount之后，在tmp目录下找到的文件列表如下。
/tmp/spark-local-20140919092949-14cc/10/shuffle_0_1_3
/tmp/spark-local-20140919092949-14cc/0f/shuffle_0_1_2
/tmp/spark-local-20140919092949-14cc/0f/shuffle_0_0_3
/tmp/spark-local-20140919092949-14cc/0c/shuffle_0_0_0
/tmp/spark-local-20140919092949-14cc/0d/shuffle_0_1_0
/tmp/spark-local-20140919092949-14cc/0d/shuffle_0_0_1
/tmp/spark-local-20140919092949-14cc/0e/shuffle_0_1_1
/tmp/spark-local-20140919092949-14cc/0e/shuffle_0_0_2
两者生成的文件数量差异非常大，具体数值计算如下
- 在HASH模式下，每一次shuffle会生成M*R的数量的文件，如上述wordcount例子中，整个job有一次shuffle过程，由于输入文件默认分片为2,故M个数为2,而spark.default.parallelism配置的值为4，故R为4,所以总共生成1*2*4=8个文件。shuffle_0_1_2解读为shuffle+shuffle_id+map_id+reduce_id，故0_1_2表示由第0次shuffle中的第1个maptask生成的文件，该文件内容会被第2个reduce task消费
- 在SORT模式下，一个Map Task只生成一个文件，而不管生成的文件要被多少的Reduce消费，故文件个数是M的数量，由于wordcount中的默认分片为2,故只生成两个data文件
多次shuffle
刚才的示例中只有一次shuffle过程，我们可以通过小小的改动来达到两次shuffle，代码如下
sc.textFile("README.md").flatMap(l => l.split(" ")).map(w => (w,1)).reduceByKey(_ + _).map(p=>(p._2,p._1)).groupByKey.collect
上述代码将reduceByKey的结果通过map进行反转，即将原来的(w, count)转换为(count,w)，然后根据出现次数进行归类。 groupByKey会再次导致数据shuffle过程。
在HASH模式下产生的文件如下所示
/tmp/spark-local-20140919094531-1cb6/12/shuffle_0_3_3
/tmp/spark-local-20140919094531-1cb6/0c/shuffle_0_0_0
/tmp/spark-local-20140919094531-1cb6/11/shuffle_0_2_3
/tmp/spark-local-20140919094531-1cb6/11/shuffle_0_3_2
/tmp/spark-local-20140919094531-1cb6/11/shuffle_1_1_3
/tmp/spark-local-20140919094531-1cb6/10/shuffle_0_2_2
/tmp/spark-local-20140919094531-1cb6/10/shuffle_0_1_3
/tmp/spark-local-20140919094531-1cb6/10/shuffle_0_3_1
/tmp/spark-local-20140919094531-1cb6/10/shuffle_1_0_3
/tmp/spark-local-20140919094531-1cb6/10/shuffle_1_1_2
/tmp/spark-local-20140919094531-1cb6/0f/shuffle_0_0_3
/tmp/spark-local-20140919094531-1cb6/0f/shuffle_0_3_0
/tmp/spark-local-20140919094531-1cb6/0f/shuffle_0_2_1
/tmp/spark-local-20140919094531-1cb6/0f/shuffle_0_1_2
/tmp/spark-local-20140919094531-1cb6/0f/shuffle_1_0_2
/tmp/spark-local-20140919094531-1cb6/0f/shuffle_1_1_1
/tmp/spark-local-20140919094531-1cb6/0d/shuffle_0_0_1
/tmp/spark-local-20140919094531-1cb6/0d/shuffle_0_1_0
/tmp/spark-local-20140919094531-1cb6/0d/shuffle_1_0_0
/tmp/spark-local-20140919094531-1cb6/0e/shuffle_0_2_0
/tmp/spark-local-20140919094531-1cb6/0e/shuffle_0_1_1
/tmp/spark-local-20140919094531-1cb6/0e/shuffle_0_0_2
/tmp/spark-local-20140919094531-1cb6/0e/shuffle_1_0_1
/tmp/spark-local-20140919094531-1cb6/0e/shuffle_1_1_0
引入一次新的shuffle，产生了大量的中间文件
如果是使用SORT，效果如何呢？只会增加M个文件，由于在新的shuffle过程中，map task数目为4,所以总共的文件是2+4=6。
/tmp/spark-local-20140919094731-034a/29/shuffle_0_3_0.data
/tmp/spark-local-20140919094731-034a/30/shuffle_0_0_0.index
/tmp/spark-local-20140919094731-034a/15/shuffle_0_1_0.data
/tmp/spark-local-20140919094731-034a/36/shuffle_0_2_0.data
/tmp/spark-local-20140919094731-034a/0c/shuffle_0_0_0.data
/tmp/spark-local-20140919094731-034a/32/shuffle_0_2_0.index
/tmp/spark-local-20140919094731-034a/32/shuffle_1_1_0.index
/tmp/spark-local-20140919094731-034a/0f/shuffle_0_1_0.index
/tmp/spark-local-20140919094731-034a/0f/shuffle_1_0_0.index
/tmp/spark-local-20140919094731-034a/0a/shuffle_1_1_0.data
/tmp/spark-local-20140919094731-034a/2b/shuffle_1_0_0.data
/tmp/spark-local-20140919094731-034a/0d/shuffle_0_3_0.index
值得指出的是shuffle_0和shuffle_1的执行次序问题，数字越大越先执行，由于spark job提交的时候是从后往前倒推的，故0是最后将执行，而前面的先执行。
Sort-based Shuffle的设计思想
sort-based shuffle的总体指导思想是一个map task最终只生成一个shuffle文件，那么后续的reduce task是如何从这一个shuffle文件中得到自己的partition呢，这个时候就需要引入一个新的文件类型即index文件。
其具体实现步骤如下:
- Map Task在读取自己输入的partition之后，将计算结果写入到ExternalSorter
- ExternalSorter会使用一个map来存储新的计算结果，新的计算结果根据partiton分类，如果是有combine操作，则需要将新的值与原有的值进行合并
- 如果ExternalSorter中的map占用的内存已经超越了使用的阀值，则将map中的内容spill到磁盘中，每一次spill产生一个不同的文件
- 当输入Partition中的所有数据都已经处理完毕之后，这时有可能一部分计算结果在内存中，另一部分计算结果在spill的一到多个文件之中，这时通过merge操作将内存和spill文件中的内容合并整到一个文件里
- 最后将每一个partition的在data文件中的起始位置和结束位置写入到index文件
相应的源文件
- SortShuffleManager.scala
- SortShuffleWriter.scala
- ExternalSorter.scala
- IndexShuffleBlockManager.scala
几个重要的函数
SortShuffleWriter.write
override def write(records: Iterator[_ >: Product2[K, V]]): Unit = {
if (dep.mapSideCombine) {
if (!dep.aggregator.isDefined) {
throw new IllegalStateException("Aggregator is empty for map-side combine")
}
sorter = new ExternalSorter[K, V, C](
dep.aggregator, Some(dep.partitioner), dep.keyOrdering, dep.serializer)
sorter.insertAll(records)
} else {
sorter = new ExternalSorter[K, V, V](
None, Some(dep.partitioner), None, dep.serializer)
sorter.insertAll(records)
}
val outputFile = shuffleBlockManager.getDataFile(dep.shuffleId, mapId)
val blockId = shuffleBlockManager.consolidateId(dep.shuffleId, mapId)
val partitionLengths = sorter.writePartitionedFile(blockId, context, outputFile)
shuffleBlockManager.writeIndexFile(dep.shuffleId, mapId, partitionLengths)
mapStatus = new MapStatus(blockManager.blockManagerId,
partitionLengths.map(MapOutputTracker.compressSize))
}
ExternalSorter.insertAll
def insertAll(records: Iterator[_ {
if (hadValue) mergeValue(oldValue, kv._2) else createCombiner(kv._2)
}
while (records.hasNext) {
elementsRead += 1
kv = records.next()
map.changeValue((getPartition(kv._1), kv._1), update)
maybeSpill(usingMap = true)
}
} else {
while (records.hasNext) {
elementsRead += 1
val kv = records.next()
buffer.insert((getPartition(kv._1), kv._1), kv._2.asInstanceOf[C])
maybeSpill(usingMap = false)
}
}
}
writePartitionedFile将内存中的数据和spill文件中内容一起合并到一个文件当中
def writePartitionedFile(
blockId: BlockId,
context: TaskContext,
outputFile: File): Array[Long] = {
val lengths = new Array[Long](numPartitions)
if (bypassMergeSort && partitionWriters != null) {
spillToPartitionFiles(if (aggregator.isDefined) map else buffer)
partitionWriters.foreach(_.commitAndClose())
var out: FileOutputStream = null
var in: FileInputStream = null
try {
out = new FileOutputStream(outputFile)
for (i <- 0 until numPartitions) {
in = new FileInputStream(partitionWriters(i).fileSegment().file)
val size = org.apache.spark.util.Utils.copyStream(in, out, false)
in.close()
in = null
lengths(i) = size
}
} finally {
if (out != null) {
out.close()
}
if (in != null) {
in.close()
}
}
} else {
for ((id, elements) <- this.partitionedIterator) {
if (elements.hasNext) {
val writer = blockManager.getDiskWriter(
blockId, outputFile, ser, fileBufferSize, context.taskMetrics.shuffleWriteMetrics.get)
for (elem
而数据读取过程中则需要使用IndexShuffleBlockManager来获取Partiton的具体位置
override def getBlockData(blockId: ShuffleBlockId): ManagedBuffer = {
val indexFile = getIndexFile(blockId.shuffleId, blockId.mapId)
val in = new DataInputStream(new FileInputStream(indexFile))
try {
in.skip(blockId.reduceId * 8)
val offset = in.readLong()
val nextOffset = in.readLong()
new FileSegmentManagedBuffer(
getDataFile(blockId.shuffleId, blockId.mapId),
offset,
nextOffset - offset)
} finally {
in.close()
}
}
参数资料
- 详细探究spark的shuffle 实现
- spark-2045 sort-based shuffle implementation
