diff --git a/core/pom.xml b/core/pom.xml index f0c936c86a..ba0071f582 100644 --- a/core/pom.xml +++ b/core/pom.xml @@ -48,6 +48,10 @@ com.ning compress-lzf + + org.xerial.snappy + snappy-java + org.ow2.asm asm diff --git a/core/src/main/scala/spark/broadcast/HttpBroadcast.scala b/core/src/main/scala/spark/broadcast/HttpBroadcast.scala index c565876950..138a8c21bc 100644 --- a/core/src/main/scala/spark/broadcast/HttpBroadcast.scala +++ b/core/src/main/scala/spark/broadcast/HttpBroadcast.scala @@ -17,21 +17,20 @@ package spark.broadcast -import com.ning.compress.lzf.{LZFInputStream, LZFOutputStream} - -import java.io._ -import java.net._ -import java.util.UUID +import java.io.{File, FileOutputStream, ObjectInputStream, OutputStream} +import java.net.URL import it.unimi.dsi.fastutil.io.FastBufferedInputStream import it.unimi.dsi.fastutil.io.FastBufferedOutputStream -import spark._ +import spark.{HttpServer, Logging, SparkEnv, Utils} +import spark.io.CompressionCodec import spark.storage.StorageLevel -import util.{MetadataCleaner, TimeStampedHashSet} +import spark.util.{MetadataCleaner, TimeStampedHashSet} + private[spark] class HttpBroadcast[T](@transient var value_ : T, isLocal: Boolean, id: Long) -extends Broadcast[T](id) with Logging with Serializable { + extends Broadcast[T](id) with Logging with Serializable { def value = value_ @@ -85,6 +84,7 @@ private object HttpBroadcast extends Logging { private val files = new TimeStampedHashSet[String] private val cleaner = new MetadataCleaner("HttpBroadcast", cleanup) + private lazy val compressionCodec = CompressionCodec.createCodec() def initialize(isDriver: Boolean) { synchronized { @@ -122,10 +122,12 @@ private object HttpBroadcast extends Logging { def write(id: Long, value: Any) { val file = new File(broadcastDir, "broadcast-" + id) - val out: OutputStream = if (compress) { - new LZFOutputStream(new FileOutputStream(file)) // Does its own buffering - } else { - new FastBufferedOutputStream(new FileOutputStream(file), bufferSize) + val out: OutputStream = { + if (compress) { + compressionCodec.compressedOutputStream(new FileOutputStream(file)) + } else { + new FastBufferedOutputStream(new FileOutputStream(file), bufferSize) + } } val ser = SparkEnv.get.serializer.newInstance() val serOut = ser.serializeStream(out) @@ -136,10 +138,12 @@ private object HttpBroadcast extends Logging { def read[T](id: Long): T = { val url = serverUri + "/broadcast-" + id - var in = if (compress) { - new LZFInputStream(new URL(url).openStream()) // Does its own buffering - } else { - new FastBufferedInputStream(new URL(url).openStream(), bufferSize) + val in = { + if (compress) { + compressionCodec.compressedInputStream(new URL(url).openStream()) + } else { + new FastBufferedInputStream(new URL(url).openStream(), bufferSize) + } } val ser = SparkEnv.get.serializer.newInstance() val serIn = ser.deserializeStream(in) diff --git a/core/src/main/scala/spark/io/CompressionCodec.scala b/core/src/main/scala/spark/io/CompressionCodec.scala new file mode 100644 index 0000000000..0adebecadb --- /dev/null +++ b/core/src/main/scala/spark/io/CompressionCodec.scala @@ -0,0 +1,82 @@ +/* + * 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 spark.io + +import java.io.{InputStream, OutputStream} + +import com.ning.compress.lzf.{LZFInputStream, LZFOutputStream} + +import org.xerial.snappy.{SnappyInputStream, SnappyOutputStream} + + +/** + * CompressionCodec allows the customization of choosing different compression implementations + * to be used in block storage. + */ +trait CompressionCodec { + + def compressedOutputStream(s: OutputStream): OutputStream + + def compressedInputStream(s: InputStream): InputStream +} + + +private[spark] object CompressionCodec { + + def createCodec(): CompressionCodec = { + // Set the default codec to Snappy since the LZF implementation initializes a pretty large + // buffer for every stream, which results in a lot of memory overhead when the number of + // shuffle reduce buckets are large. + createCodec(classOf[SnappyCompressionCodec].getName) + } + + def createCodec(codecName: String): CompressionCodec = { + Class.forName( + System.getProperty("spark.io.compression.codec", codecName), + true, + Thread.currentThread.getContextClassLoader).newInstance().asInstanceOf[CompressionCodec] + } +} + + +/** + * LZF implementation of [[spark.io.CompressionCodec]]. + */ +class LZFCompressionCodec extends CompressionCodec { + + override def compressedOutputStream(s: OutputStream): OutputStream = { + new LZFOutputStream(s).setFinishBlockOnFlush(true) + } + + override def compressedInputStream(s: InputStream): InputStream = new LZFInputStream(s) +} + + +/** + * Snappy implementation of [[spark.io.CompressionCodec]]. + * Block size can be configured by spark.io.compression.snappy.block.size. + */ +class SnappyCompressionCodec extends CompressionCodec { + + override def compressedOutputStream(s: OutputStream): OutputStream = { + val blockSize = System.getProperty("spark.io.compression.snappy.block.size", "32768").toInt + new SnappyOutputStream(s, blockSize) + } + + override def compressedInputStream(s: InputStream): InputStream = new SnappyInputStream(s) +} diff --git a/core/src/main/scala/spark/scheduler/ShuffleMapTask.scala b/core/src/main/scala/spark/scheduler/ShuffleMapTask.scala index 1c25605f75..e3bb6d1e60 100644 --- a/core/src/main/scala/spark/scheduler/ShuffleMapTask.scala +++ b/core/src/main/scala/spark/scheduler/ShuffleMapTask.scala @@ -18,16 +18,9 @@ package spark.scheduler import java.io._ -import java.util.{HashMap => JHashMap} import java.util.zip.{GZIPInputStream, GZIPOutputStream} -import scala.collection.mutable.{ArrayBuffer, HashMap} -import scala.collection.JavaConversions._ - -import it.unimi.dsi.fastutil.io.FastBufferedOutputStream - -import com.ning.compress.lzf.LZFInputStream -import com.ning.compress.lzf.LZFOutputStream +import scala.collection.mutable.HashMap import spark._ import spark.executor.ShuffleWriteMetrics @@ -109,11 +102,7 @@ private[spark] class ShuffleMapTask( preferredLocs.foreach (hostPort => Utils.checkHost(Utils.parseHostPort(hostPort)._1, "preferredLocs : " + preferredLocs)) } - var split = if (rdd == null) { - null - } else { - rdd.partitions(partition) - } + var split = if (rdd == null) null else rdd.partitions(partition) override def writeExternal(out: ObjectOutput) { RDDCheckpointData.synchronized { diff --git a/core/src/main/scala/spark/storage/BlockManager.scala b/core/src/main/scala/spark/storage/BlockManager.scala index e4ffa57ad2..3a72474419 100644 --- a/core/src/main/scala/spark/storage/BlockManager.scala +++ b/core/src/main/scala/spark/storage/BlockManager.scala @@ -27,11 +27,10 @@ import akka.dispatch.{Await, Future} import akka.util.Duration import akka.util.duration._ -import com.ning.compress.lzf.{LZFInputStream, LZFOutputStream} - import it.unimi.dsi.fastutil.io.FastByteArrayOutputStream import spark.{Logging, SparkEnv, SparkException, Utils} +import spark.io.CompressionCodec import spark.network._ import spark.serializer.Serializer import spark.util.{ByteBufferInputStream, IdGenerator, MetadataCleaner, TimeStampedHashMap} @@ -158,6 +157,13 @@ private[spark] class BlockManager( val metadataCleaner = new MetadataCleaner("BlockManager", this.dropOldBlocks) initialize() + // The compression codec to use. Note that the "lazy" val is necessary because we want to delay + // the initialization of the compression codec until it is first used. The reason is that a Spark + // program could be using a user-defined codec in a third party jar, which is loaded in + // Executor.updateDependencies. When the BlockManager is initialized, user level jars hasn't been + // loaded yet. + private lazy val compressionCodec: CompressionCodec = CompressionCodec.createCodec() + /** * Construct a BlockManager with a memory limit set based on system properties. */ @@ -919,18 +925,14 @@ private[spark] class BlockManager( * Wrap an output stream for compression if block compression is enabled for its block type */ def wrapForCompression(blockId: String, s: OutputStream): OutputStream = { - if (shouldCompress(blockId)) { - (new LZFOutputStream(s)).setFinishBlockOnFlush(true) - } else { - s - } + if (shouldCompress(blockId)) compressionCodec.compressedOutputStream(s) else s } /** * Wrap an input stream for compression if block compression is enabled for its block type */ def wrapForCompression(blockId: String, s: InputStream): InputStream = { - if (shouldCompress(blockId)) new LZFInputStream(s) else s + if (shouldCompress(blockId)) compressionCodec.compressedInputStream(s) else s } def dataSerialize( diff --git a/core/src/main/scala/spark/storage/DiskStore.scala b/core/src/main/scala/spark/storage/DiskStore.scala index 3495d653bd..3ebfe173b1 100644 --- a/core/src/main/scala/spark/storage/DiskStore.scala +++ b/core/src/main/scala/spark/storage/DiskStore.scala @@ -66,7 +66,6 @@ private class DiskStore(blockManager: BlockManager, rootDirs: String) override def close() { if (initialized) { objOut.close() - bs.close() channel = null bs = null objOut = null diff --git a/core/src/test/scala/spark/io/CompressionCodecSuite.scala b/core/src/test/scala/spark/io/CompressionCodecSuite.scala new file mode 100644 index 0000000000..1ba82fe2b9 --- /dev/null +++ b/core/src/test/scala/spark/io/CompressionCodecSuite.scala @@ -0,0 +1,62 @@ +/* + * 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 spark.io + +import java.io.{ByteArrayInputStream, ByteArrayOutputStream} + +import org.scalatest.FunSuite + + +class CompressionCodecSuite extends FunSuite { + + def testCodec(codec: CompressionCodec) { + // Write 1000 integers to the output stream, compressed. + val outputStream = new ByteArrayOutputStream() + val out = codec.compressedOutputStream(outputStream) + for (i <- 1 until 1000) { + out.write(i % 256) + } + out.close() + + // Read the 1000 integers back. + val inputStream = new ByteArrayInputStream(outputStream.toByteArray) + val in = codec.compressedInputStream(inputStream) + for (i <- 1 until 1000) { + assert(in.read() === i % 256) + } + in.close() + } + + test("default compression codec") { + val codec = CompressionCodec.createCodec() + assert(codec.getClass === classOf[SnappyCompressionCodec]) + testCodec(codec) + } + + test("lzf compression codec") { + val codec = CompressionCodec.createCodec(classOf[LZFCompressionCodec].getName) + assert(codec.getClass === classOf[LZFCompressionCodec]) + testCodec(codec) + } + + test("snappy compression codec") { + val codec = CompressionCodec.createCodec(classOf[SnappyCompressionCodec].getName) + assert(codec.getClass === classOf[SnappyCompressionCodec]) + testCodec(codec) + } +} diff --git a/docs/configuration.md b/docs/configuration.md index 5c06897cae..0bcd73ca99 100644 --- a/docs/configuration.md +++ b/docs/configuration.md @@ -35,7 +35,7 @@ for these variables. * `SPARK_JAVA_OPTS`, to add JVM options. This includes any system properties that you'd like to pass with `-D`. * `SPARK_CLASSPATH`, to add elements to Spark's classpath. * `SPARK_LIBRARY_PATH`, to add search directories for native libraries. -* `SPARK_MEM`, to set the amount of memory used per node. This should be in the same format as the +* `SPARK_MEM`, to set the amount of memory used per node. This should be in the same format as the JVM's -Xmx option, e.g. `300m` or `1g`. Note that this option will soon be deprecated in favor of the `spark.executor.memory` system property, so we recommend using that in new code. @@ -77,7 +77,7 @@ there are at least five properties that you will commonly want to control: Class to use for serializing objects that will be sent over the network or need to be cached in serialized form. The default of Java serialization works with any Serializable Java object but is quite slow, so we recommend using spark.KryoSerializer - and configuring Kryo serialization when speed is necessary. Can be any subclass of + and configuring Kryo serialization when speed is necessary. Can be any subclass of spark.Serializer). @@ -86,7 +86,7 @@ there are at least five properties that you will commonly want to control: (none) If you use Kryo serialization, set this class to register your custom classes with Kryo. - You need to set it to a class that extends + You need to set it to a class that extends spark.KryoRegistrator). See the tuning guide for more details. @@ -180,6 +180,21 @@ Apart from these, the following properties are also available, and may be useful Can save substantial space at the cost of some extra CPU time. + + spark.io.compression.codec + spark.io.SnappyCompressionCodec + + The compression codec class to use for various compressions. By default, Spark provides two + codecs: spark.io.LZFCompressionCodec and spark.io.SnappyCompressionCodec. + + + + spark.io.compression.snappy.block.size + 32768 + + Block size (in bytes) used in Snappy compression, in the case when Snappy compression codec is used. + + spark.reducer.maxMbInFlight 48 diff --git a/pom.xml b/pom.xml index 44729bd422..f3bca6c40b 100644 --- a/pom.xml +++ b/pom.xml @@ -182,6 +182,11 @@ compress-lzf 0.8.4 + + org.xerial.snappy + snappy-java + 1.0.5 + org.ow2.asm asm diff --git a/project/SparkBuild.scala b/project/SparkBuild.scala index 9920e00a67..07d7b76901 100644 --- a/project/SparkBuild.scala +++ b/project/SparkBuild.scala @@ -151,6 +151,7 @@ object SparkBuild extends Build { val excludeJackson = ExclusionRule(organization = "org.codehaus.jackson") val excludeNetty = ExclusionRule(organization = "org.jboss.netty") val excludeAsm = ExclusionRule(organization = "asm") + val excludeSnappy = ExclusionRule(organization = "org.xerial.snappy") def coreSettings = sharedSettings ++ Seq( name := "spark-core", @@ -168,6 +169,7 @@ object SparkBuild extends Build { "org.slf4j" % "slf4j-log4j12" % slf4jVersion, "commons-daemon" % "commons-daemon" % "1.0.10", "com.ning" % "compress-lzf" % "0.8.4", + "org.xerial.snappy" % "snappy-java" % "1.0.5", "org.ow2.asm" % "asm" % "4.0", "com.google.protobuf" % "protobuf-java" % "2.4.1", "com.typesafe.akka" % "akka-actor" % "2.0.5" excludeAll(excludeNetty), @@ -235,6 +237,7 @@ object SparkBuild extends Build { exclude("jline","jline") exclude("log4j","log4j") exclude("org.apache.cassandra.deps", "avro") + excludeAll(excludeSnappy) ) ) @@ -257,7 +260,7 @@ object SparkBuild extends Build { "Akka Repository" at "http://repo.akka.io/releases/" ), libraryDependencies ++= Seq( - "org.apache.flume" % "flume-ng-sdk" % "1.2.0" % "compile" excludeAll(excludeNetty), + "org.apache.flume" % "flume-ng-sdk" % "1.2.0" % "compile" excludeAll(excludeNetty, excludeSnappy), "com.github.sgroschupf" % "zkclient" % "0.1" excludeAll(excludeNetty), "org.twitter4j" % "twitter4j-stream" % "3.0.3" excludeAll(excludeNetty), "com.typesafe.akka" % "akka-zeromq" % "2.0.5" excludeAll(excludeNetty) diff --git a/streaming/src/main/scala/spark/streaming/Checkpoint.scala b/streaming/src/main/scala/spark/streaming/Checkpoint.scala index 1e4c1e3742..070d930b5e 100644 --- a/streaming/src/main/scala/spark/streaming/Checkpoint.scala +++ b/streaming/src/main/scala/spark/streaming/Checkpoint.scala @@ -17,16 +17,17 @@ package spark.streaming -import spark.{Logging, Utils} - -import org.apache.hadoop.fs.{FileUtil, Path} -import org.apache.hadoop.conf.Configuration - import java.io._ -import com.ning.compress.lzf.{LZFInputStream, LZFOutputStream} import java.util.concurrent.Executors import java.util.concurrent.RejectedExecutionException +import org.apache.hadoop.fs.Path +import org.apache.hadoop.conf.Configuration + +import spark.Logging +import spark.io.CompressionCodec + + private[streaming] class Checkpoint(@transient ssc: StreamingContext, val checkpointTime: Time) extends Logging with Serializable { @@ -49,6 +50,7 @@ class Checkpoint(@transient ssc: StreamingContext, val checkpointTime: Time) } } + /** * Convenience class to speed up the writing of graph checkpoint to file */ @@ -66,6 +68,8 @@ class CheckpointWriter(checkpointDir: String) extends Logging { val maxAttempts = 3 val executor = Executors.newFixedThreadPool(1) + private val compressionCodec = CompressionCodec.createCodec() + // Removed code which validates whether there is only one CheckpointWriter per path 'file' since // I did not notice any errors - reintroduce it ? @@ -103,7 +107,7 @@ class CheckpointWriter(checkpointDir: String) extends Logging { def write(checkpoint: Checkpoint) { val bos = new ByteArrayOutputStream() - val zos = new LZFOutputStream(bos) + val zos = compressionCodec.compressedOutputStream(bos) val oos = new ObjectOutputStream(zos) oos.writeObject(checkpoint) oos.close() @@ -137,6 +141,8 @@ object CheckpointReader extends Logging { val fs = new Path(path).getFileSystem(new Configuration()) val attempts = Seq(new Path(path, "graph"), new Path(path, "graph.bk"), new Path(path), new Path(path + ".bk")) + val compressionCodec = CompressionCodec.createCodec() + attempts.foreach(file => { if (fs.exists(file)) { logInfo("Attempting to load checkpoint from file '" + file + "'") @@ -147,7 +153,7 @@ object CheckpointReader extends Logging { // of ObjectInputStream is used to explicitly use the current thread's default class // loader to find and load classes. This is a well know Java issue and has popped up // in other places (e.g., http://jira.codehaus.org/browse/GROOVY-1627) - val zis = new LZFInputStream(fis) + val zis = compressionCodec.compressedInputStream(fis) val ois = new ObjectInputStreamWithLoader(zis, Thread.currentThread().getContextClassLoader) val cp = ois.readObject.asInstanceOf[Checkpoint] ois.close() @@ -170,7 +176,9 @@ object CheckpointReader extends Logging { } private[streaming] -class ObjectInputStreamWithLoader(inputStream_ : InputStream, loader: ClassLoader) extends ObjectInputStream(inputStream_) { +class ObjectInputStreamWithLoader(inputStream_ : InputStream, loader: ClassLoader) + extends ObjectInputStream(inputStream_) { + override def resolveClass(desc: ObjectStreamClass): Class[_] = { try { return loader.loadClass(desc.getName())