akka actor模型和java thread模型各有啥优势

Posted 2023-04-23

　　1) akka actor背后也是通过线程池来执行的。actor与thread不是一个维度的事情，actor比thread更抽象，更偏上层，它只是定义了一种机制，怎么实现或底层用什么实现是另一回事。
2) 实际上actor更容易驾驭(框架帮你屏蔽了很多底层的事情)，thread在并发下兼顾性能和正确性有很多技巧，要用好还没那么容易。 参考技术A 了解java内存模型对深入了解jvm有很多好处。JMM通过控制主内存与每个线程的本地内存之间的交互，来为开发者提供内存可见性保证。

Scala笔记整理：Actor和AKKA

[TOC]

---

概述

? Scala的Actor有点类似于Java中的多线程编程。但是不同的是，Scala的Actor提供的模型与多线程有所不同。Scala的Actor尽可能地避免锁和共享状态，从而避免多线程并发时出现资源争用的情况，进而提升多线程编程的性能。

Spark中使用的分布式多线程框架，是Akka，是Scala的一种多线程的类库。Akka也实现了类似Scala Actor的模型，其核心概念同样也是Actor。Scala Actor模型已经在2.1.0的时候还在用，但是在2.1.1的时候已经被遗弃了，Spark开始转换用AKKA来替代Scala Actor，但是Scala Actor概念和原理都还是相同的。所以学习Scala Actor对我们学习AKKA，Spark还是有所帮助的

之所以学习Scala Actor,AKKA是因为在学习Spark源码的时，我们能看懂Spark的源码，因为在底层消息传递机制上大量使用AKKA的传送机制。

scala actor

在使用前，需要先引入maven依赖：

<!--scala actor-->
<dependency>
    <groupId>org.scala-lang</groupId>
    <artifactId>scala-actors</artifactId>
    <version>2.10.5</version>
</dependency>

actor单向通信

测试代码如下：

package cn.xpleaf.bigdata.p5.myactor
import scala.actors.Actor
/**
  * 学习scala actor的基本操作
  * 和java中的Runnable Thread几乎一致
  *
  * 第一步：编写一个类，扩展特质trait Actor（scala 的actor）
  * 第二步：复写其中的act方法
  * 第三步：创建该actor的对象，调用该对象的start()方法，启动该线程
  * 第四步：通过scala的操作符"!"，发送消息
  * 第五步：结束的话，调用close即可
  *
  * 模拟单向打招呼
  */
object ActorOps {
    def main(args: Array[String]): Unit = {
        val mFActor = new MyFirstActor()
        mFActor.start()
        // 发送消息
        mFActor ! "小美，睡了吗？"
        mFActor ! "我去洗澡了~"
        mFActor ! "呵呵"
    }
}

class MyFirstActor extends Actor {
    override def act(): Unit = {
        while(true) {
            receive {
                case str: String => println(str)
            }
        }
    }
}

输出结果如下：

小美，睡了吗？
我去洗澡了~
呵呵

使用样例类（case class）进行actor消息传递

测试代码如下：

package cn.xpleaf.bigdata.p5.myactor

import scala.actors.Actor

/**
  *
  */
object GreetingActor {
    def main(args: Array[String]): Unit = {
        val ga = new GreetingActor
        ga.start()

        ga ! Greeting("小美")
        ga ! WorkContent("装系统")
    }
}

case class Greeting(name:String)
case class WorkContent(content:String)

class GreetingActor extends Actor {
    override def act(): Unit = {
        while(true) {
            receive {
                case Greeting(name) => println(s"Hello, $name")
                case WorkContent(content) => println(s"Let‘s talk about sth. with $content")
            }
        }
    }
}

输出结果如下：

Hello, 小美
Let‘s talk about sth. with 装系统

actor相互通信

测试代码如下：

package cn.xpleaf.bigdata.p5.myactor

import scala.actors.Actor

/**
  * actor之线程间，互相通信
  *
  * studentActor
  *     向老师问了一个问题
  *
  * teacherActor
  *     向学生做回应
  *
  * 通信的协议：
  * 请求，使用Request(内容)来表示
  * 响应，使用Response(内容)来表示
  */
object _03CommunicationActorOps {
    def main(args: Array[String]): Unit = {
        val teacherActor = new TeacherActor()
        teacherActor.start()
        val studentActor = new StudentActor(teacherActor)
        studentActor.start()

        studentActor ! Request("老李啊，scala学习为什么这么难啊")
    }
}

case class Request(req:String)
case class Response(resp:String)

class StudentActor(teacherActor: TeacherActor) extends Actor {
    override def act(): Unit = {
        while(true) {
            receive {
                case Request(req) => {
                    // 向老师请求相关的问题
                    println("学生向老师说：" + req)
                    teacherActor ! Request(req)
                }
                case Response(resp) => {
                    println(resp)
                    println("高！")
                }
            }
        }
    }
}

class TeacherActor() extends Actor {
    override def act(): Unit = {
        while (true) {
            receive {
                case Request(req) => {  // 接收到学生的请求
                    sender ! Response("这个问题，需要如此搞定~")
                }
            }
        }
    }
}

输出结果如下：

学生向老师说：老李啊，scala学习为什么这么难啊
这个问题，需要如此搞定~
高！

消息的同步和Future

1、Scala在默认情况下，消息都是以异步进行发送的；但是如果发送的消息是同步的，即对方接受后，一定要给自己返回结果，那么可以使用!?的方式发送消息。即：

val response= activeActor !? activeMessage

2、如果要异步发送一个消息，但是在后续要获得消息的返回值，那么可以使用Future。即!!语法，如下：

val futureResponse = activeActor !! activeMessage
val activeReply = future()

AKKA actor

首先需要添加akka的maven依赖：

<!--akka actor-->
<dependency>
    <groupId>com.typesafe.akka</groupId>
    <artifactId>akka-actor_2.10</artifactId>
    <version>2.3.16</version>
</dependency>
<dependency>
    <groupId>com.typesafe.akka</groupId>
    <artifactId>akka-remote_2.10</artifactId>
    <version>2.3.16</version>
</dependency>
<dependency>
    <groupId>com.typesafe.akka</groupId>
    <artifactId>akka-slf4j_2.10</artifactId>
    <version>2.3.16</version>
</dependency>

AKKA消息传递——本地

原理如下：

_01StudentActorOps

package cn.xpleaf.bigdata.p5.myakka.p1

import akka.actor.{Actor, ActorSystem, Props}
import cn.xpleaf.bigdata.p5.myakka.MessageProtocol.{QuoteRequest, QuoteResponse}

import scala.util.Random

/**
  * 基于AKKA Actor的单向通信案例
  * 学生向老师发送请求
  */
object _01StudentActorOps {
    def main(args: Array[String]): Unit = {
        // 第一步：构建Actor操作系统
        val actorSystem = ActorSystem("StudentActorSystem")
        // 第二步：actorSystem创建TeacherActor的代理对象ActorRef
        val teacherActorRef = actorSystem.actorOf(Props[TeacherActor])
        // 第三步：发送消息
        teacherActorRef ! QuoteRequest()

        Thread.sleep(2000)
        // 第四步：关闭
        actorSystem.shutdown()
    }
}

class TeacherActor extends Actor {
    val quotes = List(
        "Moderation is for cowards",
        "Anything worth doing is worth overdoing",
        "The trouble is you think you have time",
        "You never gonna know if you never even try")

    override def receive = {
        case QuoteRequest() => {
            val random = new Random()

            val randomIndex = random.nextInt(quotes.size)
            val randomQuote = quotes(randomIndex)

            val response = QuoteResponse(randomQuote)
            println(response)
        }
    }
}

MessageProtocol

后面akka通信的几个测试程序都会使用到这个object，只在这里给出，后面不再给出。

package cn.xpleaf.bigdata.p5.myakka

/**
  * akka actor通信协议
  */
object MessageProtocol {

    case class QuoteRequest()

    case class QuoteResponse(resp: String)

    case class InitSign()

}

object Start extends Serializable

object Stop extends Serializable

trait Message {
    val id: String
}

case class Shutdown(waitSecs: Int) extends Serializable

case class Heartbeat(id: String, magic: Int) extends Message with Serializable

case class Header(id: String, len: Int, encrypted: Boolean) extends Message with Serializable

case class Packet(id: String, seq: Long, content: String) extends Message with Serializable

测试

输出结果如下：

QuoteResponse(Anything worth doing is worth overdoing)

AKKA请求与响应——本地

原理如下：

TeacherActor

package cn.xpleaf.bigdata.p5.myakka.p2

import akka.actor.Actor
import cn.xpleaf.bigdata.p5.myakka.MessageProtocol.{QuoteRequest, QuoteResponse}

import scala.util.Random

/**
  * Teacher Actor
  */
class TeacherActor extends Actor {
    val quotes = List(
        "Moderation is for cowards",
        "Anything worth doing is worth overdoing",
        "The trouble is you think you have time",
        "You never gonna know if you never even try")

    override def receive = {
        case QuoteRequest() => {
            val random = new Random()

            val randomIndex = random.nextInt(quotes.size)
            val randomQuote = quotes(randomIndex)

            val response = QuoteResponse(randomQuote)
//            println(response)
            sender ! response
        }
    }
}

StudentActor

package cn.xpleaf.bigdata.p5.myakka.p2

import akka.actor.{Actor, ActorLogging, ActorRef}
import cn.xpleaf.bigdata.p5.myakka.MessageProtocol.{InitSign, QuoteRequest, QuoteResponse}

/**
  * Student Actor
  * 当学生接收到InitSign信号之后，便向老师发送一条Request请求的消息
  */
class StudentActor(teacherActorRef:ActorRef) extends Actor with ActorLogging {
    override def receive = {
        case InitSign => {
            teacherActorRef ! QuoteRequest()
//            println("student send request")
        }
        case QuoteResponse(resp) => {
            log.info(s"$resp")
        }
    }
}

DriverApp

package cn.xpleaf.bigdata.p5.myakka.p2

import akka.actor.{ActorSystem, Props}
import cn.xpleaf.bigdata.p5.myakka.MessageProtocol.InitSign

object DriverApp {
    def main(args: Array[String]): Unit = {
        val actorSystem = ActorSystem("teacherStudentSystem")
        // 老师的代理对象
        val teacherActorRef = actorSystem.actorOf(Props[TeacherActor], "teacherActor")
        // 学生的代理对象
        val studentActorRef = actorSystem.actorOf(Props[StudentActor](new StudentActor(teacherActorRef)), "studentActor")

        studentActorRef ! InitSign

        Thread.sleep(2000)

        actorSystem.shutdown()
    }
}

测试

输出结果如下：

[INFO] [04/24/2018 10:02:19.932] [teacherStudentSystem-akka.actor.default-dispatcher-2] [akka://teacherStudentSystem/user/studentActor] Anything worth doing is worth overdoing

AKKA请求与响应——远程

application.conf

MyRemoteServerSideActor {
  akka {
    actor {
      provider = "akka.remote.RemoteActorRefProvider"
    }
    remote {
      enabled-transports = ["akka.remote.netty.tcp"]
      netty.tcp {
        hostname = "127.0.0.1"
        port = 2552
      }
    }
  }
}
MyRemoteClientSideActor {
  akka {
    actor {
      provider = "akka.remote.RemoteActorRefProvider"
    }
  }
}

RemoteActor

package cn.xpleaf.bigdata.p5.myakka.p3

import akka.actor.{Actor, ActorLogging}
import cn.xpleaf.bigdata.p5.myakka.{Header, Shutdown, Start, Stop}

class RemoteActor extends Actor with ActorLogging {
    def receive = {
        case Start => { // 处理Start消息
            log.info("Remote Server Start ==>RECV Start event : " + Start)
        }
        case Stop => { // 处理Stop消息
            log.info("Remote Server Stop ==>RECV Stop event: " + Stop)
        }
        case Shutdown(waitSecs) => { // 处理Shutdown消息
            log.info("Remote Server Shutdown ==>Wait to shutdown: waitSecs=" + waitSecs)
            Thread.sleep(waitSecs)
            log.info("Remote Server Shutdown ==>Shutdown this system.")
            context.system.shutdown // 停止当前ActorSystem系统
        }
        case Header(id, len, encrypted) => log.info("Remote Server => RECV header: " + (id, len, encrypted)) // 处理Header消息
        case _ =>
    }
}

AkkaServerApplication

package cn.xpleaf.bigdata.p5.myakka.p3

import akka.actor.{ActorSystem, Props}
import com.typesafe.config.ConfigFactory

object AkkaServerApplication extends App {
    // 创建名称为remote-system的ActorSystem：从配置文件application.conf中获取该Actor的配置内容
    val system = ActorSystem("remote-system",
        ConfigFactory.load().getConfig("MyRemoteServerSideActor"))

    val log = system.log
    log.info("===>Remote server actor started: " + system)
    // 创建一个名称为remoteActor的Actor，返回一个ActorRef，这里我们不需要使用这个返回值
    system.actorOf(Props[RemoteActor], "remoteActor")

}

ClientActor

package cn.xpleaf.bigdata.p5.myakka.p3

import akka.actor.SupervisorStrategy.Stop
import akka.actor.{Actor, ActorLogging}
import cn.xpleaf.bigdata.p5.myakka.{Header, Start}

class ClientActor extends Actor with ActorLogging {

    // akka.<protocol>://<actor system>@<hostname>:<port>/<actor path>
    val path = "akka.tcp://[email protected]:2552/user/remoteActor" // 远程Actor的路径，通过该路径能够获取到远程Actor的一个引用
    val remoteServerRef = context.actorSelection(path) // 获取到远程Actor的一个引用，通过该引用可以向远程Actor发送消息
    @volatile var connected = false
    @volatile var stop = false

    def receive = {
        case Start => { // 发送Start消息表示要与远程Actor进行后续业务逻辑处理的通信，可以指示远程Actor初始化一些满足业务处理的操作或数据
            send(Start)
            if (!connected) {
                connected = true
                log.info("ClientActor==> Actor connected: " + this)
            }
        }
        case Stop => {
            send(Stop)
            stop = true
            connected = false
            log.info("ClientActor=> Stopped")
        }
        case header: Header => {
            log.info("ClientActor=> Header")
            send(header)
        }
        case (seq, result) => log.info("RESULT: seq=" + seq + ", result=" + result) // 用于接收远程Actor处理一个Packet消息的结果
        case m => log.info("Unknown message: " + m)
    }

    private def send(cmd: Serializable): Unit = {
        log.info("Send command to server: " + cmd)
        try {
            remoteServerRef ! cmd // 发送一个消息到远程Actor，消息必须是可序列化的，因为消息对象要经过网络传输
        } catch {
            case e: Exception => {
                connected = false
                log.info("Try to connect by sending Start command...")
                send(Start)
            }
        }
    }

}

AkkaClientApplication

package cn.xpleaf.bigdata.p5.myakka.p3

import akka.actor.{ActorSystem, Props}
import cn.xpleaf.bigdata.p5.myakka.{Header, Start}
import com.typesafe.config.ConfigFactory

object AkkaClientApplication extends App {
    // 通过配置文件application.conf配置创建ActorSystem系统
    val system = ActorSystem("client-system",
        ConfigFactory.load().getConfig("MyRemoteClientSideActor"))
    val log = system.log
    val clientActor = system.actorOf(Props[ClientActor], "clientActor") // 获取到ClientActor的一个引用
    clientActor ! Start // 发送一个Start消息，第一次与远程Actor握手（通过本地ClientActor进行转发）
    Thread.sleep(2000)
    clientActor ! Header("What‘s your name: Can you tell me ", 20, encrypted = false) // 发送一个Header消息到远程Actor（通过本地ClientActor进行转发）
    Thread.sleep(2000)

}

测试

服务端输出结果如下：

[INFO] [04/24/2018 09:39:49.271] [main] [Remoting] Starting remoting
[INFO] [04/24/2018 09:39:49.508] [main] [Remoting] Remoting started; listening on addresses :[akka.tcp://[email protected]:2552]
[INFO] [04/24/2018 09:39:49.509] [main] [Remoting] Remoting now listens on addresses: [akka.tcp://[email protected]:2552]
[INFO] [04/24/2018 09:39:49.517] [main] [ActorSystem(remote-system)] ===>Remote server actor started: akka://remote-system
[INFO] [04/24/2018 09:46:01.872] [remote-system-akka.actor.default-dispatcher-3] [akka.tcp://[email protected]:2552/user/remoteActor] Remote Server Start ==>RECV Start event : [email protected]
[INFO] [04/24/2018 09:46:03.501] [remote-system-akka.actor.default-dispatcher-3] [akka.tcp://[email protected]:2552/user/remoteActor] Remote Server => RECV header: (What‘s your name: Can you tell me ,20,false)

客户端输出结果如下：

[INFO] [04/24/2018 09:46:01.274] [main] [Remoting] Starting remoting
[INFO] [04/24/2018 09:46:01.479] [main] [Remoting] Remoting started; listening on addresses :[akka.tcp://[email protected]:2552]
[INFO] [04/24/2018 09:46:01.480] [main] [Remoting] Remoting now listens on addresses: [akka.tcp://[email protected]:2552]
[INFO] [04/24/2018 09:46:01.493] [client-system-akka.actor.default-dispatcher-4] [akka.tcp://[email protected]:2552/user/clientActor] Send command to server: [email protected]
[INFO] [04/24/2018 09:46:01.496] [client-system-akka.actor.default-dispatcher-4] [akka.tcp://[email protected]:2552/user/clientActor] ClientActor==> Actor connected: [email protected]
[INFO] [04/24/2018 09:46:03.490] [client-system-akka.actor.default-dispatcher-2] [akka.tcp://[email protected]:2552/user/clientActor] ClientActor=> Header
[INFO] [04/24/2018 09:46:03.491] [client-system-akka.actor.default-dispatcher-2] [akka.tcp://[email protected]:2552/user/clientActor] Send command to server: Header(What‘s your name: Can you tell me ,20,false)