延迟调度算法的实现是在TaskSetManager类中的,它通过将task存放在四个不同级别的hash表里,当有可用的资源时,resourceOffer函数的参数之一(maxLocality)就是这些资源的最大(或者最优)locality级别,如果存在task满足资源的locality,那从最优级别的hash表。也就是task和excutor都有loclity级别,如果能找到匹配的task,那从匹配的task中找一个最优的task。
  
=====================延迟调度算法=============================
->TaskSetManager::resourceOffer(execId: String, host: String,maxLocality: TaskLocality.TaskLocality): Option[TaskDescription]
->if (maxLocality != TaskLocality.NO_PREF) --如果资源是有locality特征的
->allowedLocality = getAllowedLocalityLevel(curTime) --获取当前taskSet允许执行的locality。getAllowedLocalityLevel随时间而变化
->if (allowedLocality > maxLocality)  --如果资源的locality级别高于taskSet允许的级别
->allowedLocality = maxLocality --那么提升taskSet的级别
->task =  findTask(execId, host, allowedLocality) --根据允许的locality级别去找一个满足要求的task
->从最优的locality级别(process_local)开始找,返回一个满足locolity的task(为最优级别)
->task match case Some((index, taskLocality, speculative)) --找到了一个task
-> val info = new TaskInfo(taskId, index, attemptNum, curTime, execId, host, taskLocality, speculative)
->if (maxLocality != TaskLocality.NO_PREF) // NO_PREF will not affect the variables related to delay scheduling
->currentLocalityIndex = getLocalityIndex(taskLocality) // Update our locality level for delay scheduling
->lastLaunchTime = curTime --更新最近执行task的时间,计算当前locality时需要
->addRunningTask(taskId) --加入执行task中
->logInfo("Starting %s (TID %d, %s, %s, %d bytes)"
->sched.dagScheduler.taskStarted(task, info) --通知调度器有task开始运行
->eventProcessActor ! BeginEvent(task, taskInfo)
->return Some(new TaskDescription(taskId, execId, taskName, index, serializedTask)) --返回task
->case _ => return None --没有满足locality要求的task,返回None
=====================end==================================

myLocalityLevels :记录当前所有有效的locality级别
localityWaits :记录不同locality级别的等待时间
currentLocalityIndex :当前的locality级别,随着等待时间而不断变化
pendingTasksForExecutor: PROCESS_LOCAL进程级别的task
pendingTasksForHost :NODE_LOCAL主机界别的task
pendingTasksForRack :机架级别的task
pendingTasksWithNoPrefs :没有locality要求的task
// Figure out which locality levels we have in our TaskSet, so we can do delay scheduling
var myLocalityLevels = computeValidLocalityLevels()
var localityWaits = myLocalityLevels.map(getLocalityWait) // Time to wait at each level
// Delay scheduling variables: we keep track of our current locality level and the time we
// last launched a task at that level, and move up a level when localityWaits[curLevel] expires.
// We then move down if we manage to launch a "more local" task.
var currentLocalityIndex = 0 // Index of our current locality level in validLocalityLevels
// Set of pending tasks for each executor. These collections are actually
// treated as stacks, in which new tasks are added to the end of the
// ArrayBuffer and removed from the end. This makes it faster to detect
// tasks that repeatedly fail because whenever a task failed, it is put
// back at the head of the stack. They are also only cleaned up lazily;
// when a task is launched, it remains in all the pending lists except
// the one that it was launched from, but gets removed from them later.
private val pendingTasksForExecutor = new HashMap[String, ArrayBuffer[Int]]

// Set of pending tasks for each host. Similar to pendingTasksForExecutor,
// but at host level.
private val pendingTasksForHost = new HashMap[String, ArrayBuffer[Int]]

// Set of pending tasks for each rack -- similar to the above.
private val pendingTasksForRack = new HashMap[String, ArrayBuffer[Int]]

// Set containing pending tasks with no locality preferences.
var pendingTasksWithNoPrefs = new ArrayBuffer[Int]
计算当前调度器中有效的locality级别
var lastLaunchTime = clock.getTime()  // Time we last launched a task at this level/**
* Compute the locality levels used in this TaskSet. Assumes that all tasks have already been
* added to queues using addPendingTask.
*
*/
private def computeValidLocalityLevels(): Array[TaskLocality.TaskLocality] = {
import TaskLocality.{PROCESS_LOCAL, NODE_LOCAL, NO_PREF, RACK_LOCAL, ANY}
val levels = new ArrayBuffer[TaskLocality.TaskLocality]
if (!pendingTasksForExecutor.isEmpty && getLocalityWait(PROCESS_LOCAL) != 0 &&
pendingTasksForExecutor.keySet.exists(sched.isExecutorAlive(_))) {
levels += PROCESS_LOCAL
}
if (!pendingTasksForHost.isEmpty && getLocalityWait(NODE_LOCAL) != 0 &&
pendingTasksForHost.keySet.exists(sched.hasExecutorsAliveOnHost(_))) {
levels += NODE_LOCAL
}
if (!pendingTasksWithNoPrefs.isEmpty) {
levels += NO_PREF
}
if (!pendingTasksForRack.isEmpty && getLocalityWait(RACK_LOCAL) != 0 &&
pendingTasksForRack.keySet.exists(sched.hasHostAliveOnRack(_))) {
levels += RACK_LOCAL
}
levels += ANY
logDebug("Valid locality levels for " + taskSet + ": " + levels.mkString(", "))
levels.toArray
}
获取每个locality级别的等待时间
private def getLocalityWait(level: TaskLocality.TaskLocality): Long = {
val defaultWait = conf.get("spark.locality.wait", "3000")
level match {
case TaskLocality.PROCESS_LOCAL =>
conf.get("spark.locality.wait.process", defaultWait).toLong
case TaskLocality.NODE_LOCAL =>
conf.get("spark.locality.wait.node", defaultWait).toLong
case TaskLocality.RACK_LOCAL =>
conf.get("spark.locality.wait.rack", defaultWait).toLong
case _ => 0L
}
}
locality的级别定义
@DeveloperApi
object TaskLocality extends Enumeration {
// Process local is expected to be used ONLY within TaskSetManager for now.
val PROCESS_LOCAL, NODE_LOCAL, NO_PREF, RACK_LOCAL, ANY = Value

type TaskLocality = Value

def isAllowed(constraint: TaskLocality, condition: TaskLocality): Boolean = {
condition <= constraint
}
}
根据输入的locality级别,获取一个在本taskSet有效的locality级别。因为当前taskSet可能有一些级别没有task。向低优先级的靠拢的原则。
/**
* Find the index in myLocalityLevels for a given locality. This is also designed to work with
* localities that are not in myLocalityLevels (in case we somehow get those) by returning the
* next-biggest level we have. Uses the fact that the last value in myLocalityLevels is ANY.
*/
def getLocalityIndex(locality: TaskLocality.TaskLocality): Int = {
var index = 0
while (locality > myLocalityLevels(index)) {
index += 1
}
index
}

获取当前允许的locality级别。它通过已经等待的时间和需要等待的时间做比较得到当前处于什么样的locality级别中。
/**
* Get the level we can launch tasks according to delay scheduling, based on current wait time.
*/
private def getAllowedLocalityLevel(curTime: Long): TaskLocality.TaskLocality = {
while (curTime - lastLaunchTime >= localityWaits(currentLocalityIndex) &&
currentLocalityIndex < myLocalityLevels.length - 1)
{
// Jump to the next locality level, and remove our waiting time for the current one since
// we don't want to count it again on the next one
lastLaunchTime += localityWaits(currentLocalityIndex)
currentLocalityIndex += 1
}
myLocalityLevels(currentLocalityIndex)
}
当一个task得到执行后,重新初始化locality级别
def recomputeLocality() {
val previousLocalityLevel = myLocalityLevels(currentLocalityIndex)
myLocalityLevels = computeValidLocalityLevels()
localityWaits = myLocalityLevels.map(getLocalityWait)
currentLocalityIndex = getLocalityIndex(previousLocalityLevel)
}
获取本taskSet有效的locality级别
/**
* Compute the locality levels used in this TaskSet. Assumes that all tasks have already been
* added to queues using addPendingTask.
*
*/
private def computeValidLocalityLevels(): Array[TaskLocality.TaskLocality] = {
import TaskLocality.{PROCESS_LOCAL, NODE_LOCAL, NO_PREF, RACK_LOCAL, ANY}
val levels = new ArrayBuffer[TaskLocality.TaskLocality]
if (!pendingTasksForExecutor.isEmpty && getLocalityWait(PROCESS_LOCAL) != 0 &&
pendingTasksForExecutor.keySet.exists(sched.isExecutorAlive(_))) {
levels += PROCESS_LOCAL
}
if (!pendingTasksForHost.isEmpty && getLocalityWait(NODE_LOCAL) != 0 &&
pendingTasksForHost.keySet.exists(sched.hasExecutorsAliveOnHost(_))) {
levels += NODE_LOCAL
}
if (!pendingTasksWithNoPrefs.isEmpty) {
levels += NO_PREF
}
if (!pendingTasksForRack.isEmpty && getLocalityWait(RACK_LOCAL) != 0 &&
pendingTasksForRack.keySet.exists(sched.hasHostAliveOnRack(_))) {
levels += RACK_LOCAL
}
levels += ANY
logDebug("Valid locality levels for " + taskSet + ": " + levels.mkString(", "))
levels.toArray
}
查找一个可符合locality要求的task。从最优的locality开始找。所以最优的locality总是优先被执行。
/**
* Dequeue a pending task for a given node and return its index and locality level.
* Only search for tasks matching the given locality constraint.
*
* @return An option containing (task index within the task set, locality, is speculative?)
*/
private def findTask(execId: String, host: String, maxLocality: TaskLocality.Value)
: Option[(Int, TaskLocality.Value, Boolean)] =
{
for (index <- findTaskFromList(execId, getPendingTasksForExecutor(execId))) {
return Some((index, TaskLocality.PROCESS_LOCAL, false))
}
。。。
// find a speculative task if all others tasks have been scheduled
findSpeculativeTask(execId, host, maxLocality).map {
case (taskIndex, allowedLocality) => (taskIndex, allowedLocality, true)}
}


spark 笔记 14: spark中的delay scheduling实现的更多相关文章

  1. spark 笔记 3:Delay Scheduling: A Simple Technique for Achieving Locality and Fairness in Cluster Scheduling

    spark论文中说他使用了延迟调度算法,源于这篇论文:http://people.csail.mit.edu/matei/papers/2010/eurosys_delay_scheduling.pd ...

  2. spark 笔记 2: Resilient Distributed Datasets: A Fault-Tolerant Abstraction for In-Memory Cluster Computing

    http://www.cs.berkeley.edu/~matei/papers/2012/nsdi_spark.pdf  ucb关于spark的论文,对spark中核心组件RDD最原始.本质的理解, ...

  3. Apache Spark 2.2.0 中文文档 - Spark RDD(Resilient Distributed Datasets)论文 | ApacheCN

    Spark RDD(Resilient Distributed Datasets)论文 概要 1: 介绍 2: Resilient Distributed Datasets(RDDs) 2.1 RDD ...

  4. spark学习笔记总结-spark入门资料精化

    Spark学习笔记 Spark简介 spark 可以很容易和yarn结合,直接调用HDFS.Hbase上面的数据,和hadoop结合.配置很容易. spark发展迅猛,框架比hadoop更加灵活实用. ...

  5. Apache Spark 2.2.0 中文文档 - Spark Streaming 编程指南 | ApacheCN

    Spark Streaming 编程指南 概述 一个入门示例 基础概念 依赖 初始化 StreamingContext Discretized Streams (DStreams)(离散化流) Inp ...

  6. Apache Spark 2.2.0 中文文档

    Apache Spark 2.2.0 中文文档 - 快速入门 | ApacheCN Geekhoo 关注 2017.09.20 13:55* 字数 2062 阅读 13评论 0喜欢 1 快速入门 使用 ...

  7. spark 笔记 7: DAGScheduler

    在前面的sparkContex和RDD都可以看到,真正的计算工作都是同过调用DAGScheduler的runjob方法来实现的.这是一个很重要的类.在看这个类实现之前,需要对actor模式有一点了解: ...

  8. Apache Spark 2.2.0 中文文档 - Spark Streaming 编程指南

    Spark Streaming 编程指南 概述 一个入门示例 基础概念 依赖 初始化 StreamingContext Discretized Streams (DStreams)(离散化流) Inp ...

  9. 二、spark入门之spark shell:文本中发现5个最常用的word

    scala> val textFile = sc.textFile("/Users/admin/spark-1.5.1-bin-hadoop2.4/README.md") s ...

随机推荐

  1. MySQL主从延时这么长,要怎么优化?

    MySQL主从复制,读写分离是互联网常见的数据库架构,该架构最令人诟病的地方就是,在数据量较大并发量较大的场景下,主从延时会比较严重. 为什么主从延时这么大? 答:MySQL使用单线程重放RelayL ...

  2. vue-router实现原理

    vue-router实现原理 近期面试,遇到关于vue-router实现原理的问题,在查阅了相关资料后,根据自己理解,来记录下.我们知道vue-router是vue的核心插件,而当前vue项目一般都是 ...

  3. [转自SA]浅谈nginx的工作原理和使用

    nginx apache 简单对比 nginx 相对 apache 的优点: 轻量级,同样起web 服务,比apache 占用更少的内存及资源 抗并发,nginx 处理请求是异步非阻塞的,而 apac ...

  4. linux无界面模式安装selenium+chrome+chromedriver并成功完成脚本(亲测可用)

    环境:docker centos 7.4 能通外网 写好的selenium脚本. 具体步骤: 一:安装selenium  这是最简单的 直接利用 pip3 install selenium 二 安装c ...

  5. Apache(web服务器)与Tomcat(应用服务器)搭建集群

    web服务器:Apache.Nginx.IIS等 应用服务器:Tomcat.JBoss.Weblogic等 现在web服务器和应用服务器其实界限已经不是太清晰了,大部分的应用服务器也包含一些web服务 ...

  6. LRU算法介绍和在JAVA的实现及源码分析

    一.写随笔的原因:最近准备去朋友公司面试,他说让我看一下LRU算法,就此整理一下,方便以后的复习. 二.具体的内容: 1.简介: LRU是Least Recently Used的缩写,即最近最少使用. ...

  7. php+ajax远程加载避免重复提交

    近日在练习签到送积分功能时,发现可以在一瞬间的时候提交好多次 导致可以重复领取多次积分 除了增加请求限制之外 发现ajax提交没有限制重复提交 遂立此贴为警示 首先上表单代码 <form ons ...

  8. 使用sproxy.exe访问基于soap的webservice

    使用vc访问基于soap的webservice有多种方法,其中有一种是使用atlsoap,关于这个可以搜索sproxy.exe文章,不在这介绍(主要是我的写作能力太差).我写这个日记主要是项记录访问w ...

  9. c# HttpClient和HttpWebRequest添加Basic类型的Authentication认证

    c#项目中用到调用客户接口,basic身份认证,base64格式加密(用户名:密码)贴上代码以备后用 1.使用HttpClient实现basic身份认证 using (HttpClient clien ...

  10. 初学者如何从零学习人工智能?(AI)

    一.机器学习 有关机器学习领域的最佳介绍,请观看Coursera的Andrew Ng机器学习课程. 它解释了基本概念,并让你很好地理解最重要的算法. 有关ML算法的简要概述,查看这个TutsPlus课 ...