spark通信原理
https://github.com/apache/spark/tree/master/core/src/main/scala/org/apache/spark/network
https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/network/BlockTransferService.scala
https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/network/BlockDataManager.scala
- package org.apache.spark.network
- import scala.reflect.ClassTag
- import org.apache.spark.network.buffer.ManagedBuffer
- import org.apache.spark.storage.{BlockId, StorageLevel}
- private[spark]
- trait BlockDataManager {
- /**
- * Interface to get local block data. Throws an exception if the block cannot be found or
- * cannot be read successfully.
- */
- def getBlockData(blockId: BlockId): ManagedBuffer
- /**
- * Put the block locally, using the given storage level.
- *
- * Returns true if the block was stored and false if the put operation failed or the block
- * already existed.
- */
- def putBlockData(
- blockId: BlockId,
- data: ManagedBuffer,
- level: StorageLevel,
- classTag: ClassTag[_]): Boolean
- /**
- * Release locks acquired by [[putBlockData()]] and [[getBlockData()]].
- */
- def releaseLock(blockId: BlockId, taskAttemptId: Option[Long]): Unit
- }
- package org.apache.spark.network
- import java.io.Closeable
- import java.nio.ByteBuffer
- import scala.concurrent.{Future, Promise}
- import scala.concurrent.duration.Duration
- import scala.reflect.ClassTag
- import org.apache.spark.internal.Logging
- import org.apache.spark.network.buffer.{FileSegmentManagedBuffer, ManagedBuffer, NioManagedBuffer}
- import org.apache.spark.network.shuffle.{BlockFetchingListener, ShuffleClient, TempFileManager}
- import org.apache.spark.storage.{BlockId, StorageLevel}
- import org.apache.spark.util.ThreadUtils
- private[spark]
- abstract class BlockTransferService extends ShuffleClient with Closeable with Logging {
- /**
- * Initialize the transfer service by giving it the BlockDataManager that can be used to fetch
- * local blocks or put local blocks.
- */
- def init(blockDataManager: BlockDataManager): Unit
- /**
- * Tear down the transfer service.
- */
- def close(): Unit
- /**
- * Port number the service is listening on, available only after [[init]] is invoked.
- */
- def port: Int
- /**
- * Host name the service is listening on, available only after [[init]] is invoked.
- */
- def hostName: String
- /**
- * Fetch a sequence of blocks from a remote node asynchronously,
- * available only after [[init]] is invoked.
- *
- * Note that this API takes a sequence so the implementation can batch requests, and does not
- * return a future so the underlying implementation can invoke onBlockFetchSuccess as soon as
- * the data of a block is fetched, rather than waiting for all blocks to be fetched.
- */
- override def fetchBlocks(
- host: String,
- port: Int,
- execId: String,
- blockIds: Array[String],
- listener: BlockFetchingListener,
- tempFileManager: TempFileManager): Unit
- /**
- * Upload a single block to a remote node, available only after [[init]] is invoked.
- */
- def uploadBlock(
- hostname: String,
- port: Int,
- execId: String,
- blockId: BlockId,
- blockData: ManagedBuffer,
- level: StorageLevel,
- classTag: ClassTag[_]): Future[Unit]
- /**
- * A special case of [[fetchBlocks]], as it fetches only one block and is blocking.
- *
- * It is also only available after [[init]] is invoked.
- */
- def fetchBlockSync(
- host: String,
- port: Int,
- execId: String,
- blockId: String,
- tempFileManager: TempFileManager): ManagedBuffer = {
- // A monitor for the thread to wait on.
- val result = Promise[ManagedBuffer]()
- fetchBlocks(host, port, execId, Array(blockId),
- new BlockFetchingListener {
- override def onBlockFetchFailure(blockId: String, exception: Throwable): Unit = {
- result.failure(exception)
- }
- override def onBlockFetchSuccess(blockId: String, data: ManagedBuffer): Unit = {
- data match {
- case f: FileSegmentManagedBuffer =>
- result.success(f)
- case _ =>
- val ret = ByteBuffer.allocate(data.size.toInt)
- ret.put(data.nioByteBuffer())
- ret.flip()
- result.success(new NioManagedBuffer(ret))
- }
- }
- }, tempFileManager)
- ThreadUtils.awaitResult(result.future, Duration.Inf)
- }
- /**
- * Upload a single block to a remote node, available only after [[init]] is invoked.
- *
- * This method is similar to [[uploadBlock]], except this one blocks the thread
- * until the upload finishes.
- */
- def uploadBlockSync(
- hostname: String,
- port: Int,
- execId: String,
- blockId: BlockId,
- blockData: ManagedBuffer,
- level: StorageLevel,
- classTag: ClassTag[_]): Unit = {
- val future = uploadBlock(hostname, port, execId, blockId, blockData, level, classTag)
- ThreadUtils.awaitResult(future, Duration.Inf)
- }
- }
spark通信原理的更多相关文章
- Spark Shuffle 堆外内存溢出问题与解决(Shuffle通信原理)
Spark Shuffle 堆外内存溢出问题与解决(Shuffle通信原理) http://xiguada.org/spark-shuffle-direct-buffer-oom/ 问题描述 Spar ...
- [Spark内核] 第32课:Spark Worker原理和源码剖析解密:Worker工作流程图、Worker启动Driver源码解密、Worker启动Executor源码解密等
本課主題 Spark Worker 原理 Worker 启动 Driver 源码鉴赏 Worker 启动 Executor 源码鉴赏 Worker 与 Master 的交互关系 [引言部份:你希望读者 ...
- Spark核心技术原理透视一(Spark运行原理)
在大数据领域,只有深挖数据科学领域,走在学术前沿,才能在底层算法和模型方面走在前面,从而占据领先地位. Spark的这种学术基因,使得它从一开始就在大数据领域建立了一定优势.无论是性能,还是方案的统一 ...
- spark核心原理
spark运行结构图如下: spark基本概念 应用程序(application):用户编写的spark应用程序,包含驱动程序(Driver)和分布在集群中多个节点上运行的Executor代码,在执行 ...
- Spark集群基础概念 与 spark架构原理
一.Spark集群基础概念 将DAG划分为多个stage阶段,遵循以下原则: 1.将尽可能多的窄依赖关系的RDD划为同一个stage阶段. 2.当遇到shuffle操作,就意味着上一个stage阶段结 ...
- Spark Worker原理和源码剖析解密:Worker工作流程图、Worker启动Driver源码解密、Worker启动Executor源码解密等
本课主题 Spark Worker 原理 Worker 启动 Driver 源码鉴赏 Worker 启动 Executor 源码鉴赏 Worker 与 Master 的交互关系 Spark Worke ...
- 基于web的IM软件通信原理分析
关于IM(InstantMessaging)即时通信类软件(如微信,QQ),大多数都是桌面应用程序或者native应用较为流行,而网上关于原生IM或桌面IM软件类的通信原理介绍也较多,此处不再赘述.而 ...
- Socket 通信原理(Android客户端和服务器以TCP&&UDP方式互通)
转载地址:http://blog.csdn.net/mad1989/article/details/9147661 ZERO.前言 有关通信原理内容是在网上或百科整理得到,代码部分为本人所写,如果不当 ...
- SSL 通信原理及Tomcat SSL 配置
SSL 通信原理及Tomcat SSL 双向配置 目录1 参考资料 .................................................................. ...
随机推荐
- 九度oj 题目1090:路径打印
题目描述: 给你一串路径,譬如:a\b\c a\d\e b\cst d\你把这些路径中蕴含的目录结构给画出来,子目录直接列在父目录下面,并比父目录向右缩一格,就像这样:a b c d ...
- 安卓ImageView.src设置图片拉伸、填满控件的方法
代码改变世界 安卓ImageView.src设置图片拉伸.填满控件的方法 需要给你的ImageView布局加上Android:adjustViewBounds="true"
- 2013 Asia acm Hangzhou Regional Contest 杭州现场赛
B Stealing Harry Potter's Precious 题目大意:给定一个n*m的地图,某些点可以走,某些点可以走某些点不可以走,给定一个起点,又给出了k个点k<=4,要求从起点 ...
- 算法复习——平面分治(hud1007)
题目: 问题描述 : Have you ever played quoit in a playground? Quoit is a game in which flat rings are pitch ...
- HashTable的构造函数有哪些
HashTable:在并发的环境下,使用synchronized将整张表锁住: HashTable构造函数有: public Hashtable(int initialCapacity, float ...
- bzoj 1818 Cqoi2010 内部白点 扫描线
[Cqoi2010]内部白点 Time Limit: 10 Sec Memory Limit: 64 MBSubmit: 1126 Solved: 530[Submit][Status][Disc ...
- java 数据库连接的几个步骤
Class.forName("oracle.jdbc.driver.OracleDriver"); String url = "jdbc:oracle:thin:@你的主 ...
- 洛谷 [P2148] E&G
SG函数的应用 首先每一组都是独立的,所以我们可以求出每一组的SG值异或出来. 那么怎么求每一组的SG值呢,网上的题解都是打表找规律,但其实这个规律是可以证明的 先看规律: x为奇数,y为奇数:SG= ...
- NOJ 1116 哈罗哈的大披萨 【淡蓝】 [状压dp+各种优化]
我只能说,珍爱生命,远离卡常数的题...感谢陈老师和蔡神,没有他们,,,我调一个星期都弄不出来,,,, 哈罗哈的大披萨 [淡蓝] 时间限制(普通/Java) : 1000 MS/ 3000 MS ...
- 词法分析器 /c++实现
#include<iostream> #include<string> #include<vector> #include<map> #include& ...