長文慎入-探索Java併發編程與高併發解決方案

所有示例代碼,請見/下載於
https://github.com/Wasabi1234/concurrency

#1 基本概念
##1.1 併發
同時擁有兩個或者多個線程，如果程序在單核處理器上運行多個線程將交替地換入或者換出內存,這些線程是同時“存在"的，每個線程都處於執行過程中的某個狀態，如果運行在多核處理器上,此時，程序中的每個線程都將分配到一個處理器核上，因此可以同時運行.
##1.2 高併發( High Concurrency)
互聯網分佈式系統架構設計中必須考慮的因素之一，通常是指，通過設計保證系統能夠同時並行處理很多請求.
##1.3 區別與聯繫

併發: 多個線程操作相同的資源，保證線程安全，合理使用資源
高併發:服務能同時處理很多請求，提高程序性能
#2 CPU
##2.1 CPU 多級緩存
爲什麼需要CPU cache
CPU的頻率太快了，快到主存跟不上
如此,在處理器時鐘週期內，CPU常常需要等待主存，浪費資源。所以cache的出現，是爲了緩解CPU和內存之間速度的不匹配問題(結構:cpu-> cache-> memory ).
CPU cache的意義
1) 時間局部性
如果某個數據被訪問，那麼在不久的將來它很可能被再次訪問
2) 空間局部性
如果某個數據被訪問，那麼與它相鄰的數據很快也可能被訪問
##2.2 緩存一致性(MESI)
用於保證多個 CPU cache 之間緩存共享數據的一致
M-modified被修改
該緩存行只被緩存在該 CPU 的緩存中,並且是被修改過的,與主存中數據是不一致的,需在未來某個時間點寫回主存,該時間是允許在其他CPU 讀取主存中相應的內存之前,當這裏的值被寫入主存之後,該緩存行狀態變爲 E
E-exclusive獨享
緩存行只被緩存在該 CPU 的緩存中,未被修改過,與主存中數據一致
可在任何時刻當被其他 CPU讀取該內存時變成 S 態,被修改時變爲 M態
S-shared共享
該緩存行可被多個 CPU 緩存,與主存中數據一致
I-invalid無效
亂序執行優化
處理器爲提高運算速度而做出違背代碼原有順序的優化
##併發的優勢與風險

#3 項目準備
##3.1 項目初始化

##3.2 併發模擬-Jmeter壓測

##3.3 併發模擬-代碼
###CountDownLatch

###Semaphore(信號量)

以上二者通常和線程池搭配

下面開始做併發模擬

package com.mmall.concurrency;

import com.mmall.concurrency.annoations.NotThreadSafe;
import lombok.extern.slf4j.Slf4j;

import java.util.concurrent.CountDownLatch;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.Semaphore;

/**
 * @author shishusheng
 * @date 18/4/1
 */
@Slf4j
@NotThreadSafe
public class ConcurrencyTest {

    /**
     * 請求總數
     */
    public static int clientTotal = 5000;

    /**
     * 同時併發執行的線程數
     */
    public static int threadTotal = 200;

    public static int count = 0;

    public static void main(String[] args) throws Exception {
        //定義線程池
        ExecutorService executorService = Executors.newCachedThreadPool();
        //定義信號量,給出允許併發的線程數目
        final Semaphore semaphore = new Semaphore(threadTotal);
        //統計計數結果
        final CountDownLatch countDownLatch = new CountDownLatch(clientTotal);
        //將請求放入線程池
        for (int i = 0; i < clientTotal ; i++) {
            executorService.execute(() -> {
                try {
                    //信號量的獲取
                    semaphore.acquire();
                    add();
                    //釋放
                    semaphore.release();
                } catch (Exception e) {
                    log.error("exception", e);
                }
                countDownLatch.countDown();
            });
        }
        countDownLatch.await();
        //關閉線程池
        executorService.shutdown();
        log.info("count:{}", count);
    }

    /**
     * 統計方法
     */
    private static void add() {
        count++;
    }
}

運行發現結果隨機,所以非線程安全
#4線程安全性
##4.1 線程安全性
當多個線程訪問某個類時，不管運行時環境採用何種調度方式或者這些進程將如何交替執行，並且在主調代碼中不需要任何額外的同步或協同，這個類都能表現出正確的行爲，那麼就稱這個類是線程安全的
##4.2 原子性
###4.2.1 Atomic 包

AtomicXXX:CAS,Unsafe.compareAndSwapInt
提供了互斥訪問，同一時刻只能有一個線程來對它進行操作
```
package com.mmall.concurrency.example.atomic;
```

import com.mmall.concurrency.annoations.ThreadSafe;
import lombok.extern.slf4j.Slf4j;

import java.util.concurrent.CountDownLatch;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.Semaphore;
import java.util.concurrent.atomic.AtomicLong;

/**

@author shishushengbr/>*/
@Slf4j
@ThreadSafe
public class AtomicExample2 {

/**
- 請求總數
  */
  public static int clientTotal = 5000;
/**
- 同時併發執行的線程數
  */
  public static int threadTotal = 200;
/**
- 工作內存
  */
  public static AtomicLong count = new AtomicLong(0);
public static void main(String[] args) throws Exception {
ExecutorService executorService = Executors.newCachedThreadPool();
final Semaphore semaphore = new Semaphore(threadTotal);
final CountDownLatch countDownLatch = new CountDownLatch(clientTotal);
for (int i = 0; i < clientTotal ; i++) {
executorService.execute(() -> {
try {
System.out.println();
semaphore.acquire();
add();
semaphore.release();
} catch (Exception e) {
log.error("exception", e);
}
countDownLatch.countDown();
});
}
countDownLatch.await();
executorService.shutdown();
//主內存
log.info("count:{}", count.get());
}

private static void add() {
count.incrementAndGet();
// count.getAndIncrement();
}
}

package com.mmall.concurrency.example.atomic;

import com.mmall.concurrency.annoations.ThreadSafe;
import lombok.extern.slf4j.Slf4j;
import java.util.concurrent.atomic.AtomicReference;

/**
 * @author shishusheng
 * @date 18/4/3
 */
@Slf4j
@ThreadSafe
public class AtomicExample4 {

    private static AtomicReference<Integer> count = new AtomicReference<>(0);

    public static void main(String[] args) {
        // 2
        count.compareAndSet(0, 2);
        // no
        count.compareAndSet(0, 1);
        // no
        count.compareAndSet(1, 3);
        // 4
        count.compareAndSet(2, 4);
        // no
        count.compareAndSet(3, 5); 
        log.info("count:{}", count.get());
    }
}

AtomicReference,AtomicReferenceFieldUpdater
AtomicBoolean
AtomicStampReference : CAS的 ABA 問題
###4.2.2 鎖
synchronized:依賴 JVM
修飾代碼塊:大括號括起來的代碼，作用於調用的對象
修飾方法: 整個方法，作用於調用的對象
修飾靜態方法:整個靜態方法，作用於所有對象
```
package com.mmall.concurrency.example.count;
```

import com.mmall.concurrency.annoations.ThreadSafe;
import lombok.extern.slf4j.Slf4j;

import java.util.concurrent.CountDownLatch;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.Semaphore;

/**

@author shishushengbr/>*/
@Slf4j
@ThreadSafe
public class CountExample3 {

/**
- 請求總數
  */
  public static int clientTotal = 5000;
/**
- 同時併發執行的線程數
  */
  public static int threadTotal = 200;
public static int count = 0;

public static void main(String[] args) throws Exception {
ExecutorService executorService = Executors.newCachedThreadPool();
final Semaphore semaphore = new Semaphore(threadTotal);
final CountDownLatch countDownLatch = new CountDownLatch(clientTotal);
for (int i = 0; i < clientTotal ; i++) {
executorService.execute(() -> {
try {
semaphore.acquire();
add();
semaphore.release();
} catch (Exception e) {
log.error("exception", e);
}
countDownLatch.countDown();
});
}
countDownLatch.await();
executorService.shutdown();
log.info("count:{}", count);
}

private synchronized static void add() {
count++;
}
}
```
synchronized 修正計數類方法
- 修飾類:括號括起來的部分，作用於所有對象
子類繼承父類的被 synchronized 修飾方法時,是沒有 synchronized 修飾的!!!
```

Lock: 依賴特殊的 CPU 指令,代碼實現
###4.2.3 對比

synchronized: 不可中斷鎖，適合競爭不激烈，可讀性好
Lock: 可中斷鎖，多樣化同步，競爭激烈時能維持常態
Atomic: 競爭激烈時能維持常態，比Lock性能好; 只能同步一
個值
##4.3 可見性
一個線程對主內存的修改可以及時的被其他線程觀察到
###4.3.1 導致共享變量在線程間不可見的原因
線程交叉執行
重排序結合線程交叉執行
共享變量更新後的值沒有在工作內存與主存間及時更新
###4.3.2 可見性之synchronized
JMM關於synchronized的規定
線程解鎖前，必須把共享變量的最新值刷新到主內存
線程加鎖時，將清空工作內存中共享變量的值，從而使
用共享變量時需要從主內存中重新讀取最新的值(加鎖與解鎖是同一把鎖)
###4.3.3 可見性之volatile
通過加入內存屏障和禁止重排序優化來實現
對volatile變量寫操作時，會在寫操作後加入一條store
屏障指令，將本地內存中的共享變量值刷新到主內存
對volatile變量讀操作時，會在讀操作前加入一條load
屏障指令，從主內存中讀取共享變量
volatile使用
```
volatile boolean inited = false;
```

//線程1:
context = loadContext();
inited= true;

// 線程2:
while( !inited ){
sleep();
}
doSomethingWithConfig(context)


##4.4 有序性
一個線程觀察其他線程中的指令執行順序，由於指令重排序的存在，該觀察結果一般雜亂無序

JMM允許編譯器和處理器對指令進行重排序，但是重排序過程不會影響到單線程程序的執行，卻會影響到多線程併發執行的正確性
###4.4.1 happens-before 規則
#5發佈對象
![](https://upload-images.jianshu.io/upload_images/4685968-ed313a1caed24223.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
![發佈對象](https://upload-images.jianshu.io/upload_images/4685968-b368f6fe5b350cbe.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
![對象逸出](https://upload-images.jianshu.io/upload_images/4685968-88d207fcc6bf1866.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
##5.1 安全發佈對象
![](https://upload-images.jianshu.io/upload_images/4685968-7400ab2abe1dbbfb.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
![非線程安全的懶漢模式](https://upload-images.jianshu.io/upload_images/4685968-ba18bdbe3a3c4ed1.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
![餓漢模式](https://upload-images.jianshu.io/upload_images/4685968-be2854c290143094.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
![線程安全的懶漢模式](https://upload-images.jianshu.io/upload_images/4685968-e632243a5a97281a.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)

package com.mmall.concurrency.example.singleton;

import com.mmall.concurrency.annoations.NotThreadSafe;

/**

懶漢模式 -》雙重同步鎖單例模式
單例實例在第一次使用時進行創建
@author shishushengbr/>*/
@NotThreadSafe
public class SingletonExample4 {

/**
- 私有構造函數
  */
  private SingletonExample4() {
}

// 1、memory = allocate() 分配對象的內存空間
// 2、ctorInstance() 初始化對象
// 3、instance = memory 設置instance指向剛分配的內存

// JVM和cpu優化，發生了指令重排

// 1、memory = allocate() 分配對象的內存空間
// 3、instance = memory 設置instance指向剛分配的內存
// 2、ctorInstance() 初始化對象

/**
- 單例對象
  */
  private static SingletonExample4 instance = null;
/**
- 靜態的工廠方法
- @return
  */
  public static SingletonExample4 getInstance() {
  // 雙重檢測機制 // B
  if (instance == null) {
  // 同步鎖
  synchronized (SingletonExample4.class) {
  if (instance == null) {
  // A - 3
  instance = new SingletonExample4();
  }
  }
  }
  return instance;
  }
  }
```
![](https://upload-images.jianshu.io/upload_images/4685968-823166cdf7936293.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
![](https://upload-images.jianshu.io/upload_images/4685968-9002671b71096f6c.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
#7 AQS
##7.1 介紹
![數據結構](https://upload-images.jianshu.io/upload_images/4685968-918dcaea77d556e9.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
- 使用Node實現FIFO隊列，可以用於構建鎖或者其他同步裝置的基礎框架
- 利用了一個int類型表示狀態
- 使用方法是繼承
- 子類通過繼承並通過實現它的方法管理其狀態{acquire 和release} 的方法操縱狀態
- 可以同時實現排它鎖和共享鎖模式(獨佔、共享)
同步組件
###CountDownLatch
```
  package com.mmall.concurrency.example.aqs;

import lombok.extern.slf4j.Slf4j;

import java.util.concurrent.CountDownLatch;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;

/**

@author shishushengbr/>*/
@Slf4j
public class CountDownLatchExample1 {

private final static int threadCount = 200;

public static void main(String[] args) throws Exception {

ExecutorService exec = Executors.newCachedThreadPool();

final CountDownLatch countDownLatch = new CountDownLatch(threadCount);

for (int i = 0; i < threadCount; i++) {
    final int threadNum = i;
    exec.execute(() -> {
        try {
            test(threadNum);
        } catch (Exception e) {
            log.error("exception", e);
        } finally {
            countDownLatch.countDown();
        }
    });
}
countDownLatch.await();
log.info("finish");
exec.shutdown();

}

private static void test(int threadNum) throws Exception {
Thread.sleep(100);
log.info("{}", threadNum);
Thread.sleep(100);
}
}

package com.mmall.concurrency.example.aqs;

import lombok.extern.slf4j.Slf4j;

import java.util.concurrent.CountDownLatch;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;

/**

指定時間內處理任務
@author shishusheng

*/br/>@Slf4j
public class CountDownLatchExample2 {

private final static int threadCount = 200;

public static void main(String[] args) throws Exception {

ExecutorService exec = Executors.newCachedThreadPool();

final CountDownLatch countDownLatch = new CountDownLatch(threadCount);

for (int i = 0; i < threadCount; i++) {
    final int threadNum = i;
    exec.execute(() -> {
        try {
            test(threadNum);
        } catch (Exception e) {
            log.error("exception", e);
        } finally {
            countDownLatch.countDown();
        }
    });
}
countDownLatch.await(10, TimeUnit.MILLISECONDS);
log.info("finish");
exec.shutdown();

}

private static void test(int threadNum) throws Exception {
Thread.sleep(100);
log.info("{}", threadNum);
}
}

##Semaphore用法
![](https://upload-images.jianshu.io/upload_images/4685968-e6cbcd4254c642c5.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
![](https://upload-images.jianshu.io/upload_images/4685968-dbefbf2c76ad5a2a.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
![](https://upload-images.jianshu.io/upload_images/4685968-41f5f5a5fd135804.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
##CycliBarrier

package com.mmall.concurrency.example.aqs;

import lombok.extern.slf4j.Slf4j;

import java.util.concurrent.CyclicBarrier;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;

/**

@author shishushengbr/>*/
@Slf4j
public class CyclicBarrierExample1 {

private static CyclicBarrier barrier = new CyclicBarrier(5);

public static void main(String[] args) throws Exception {

ExecutorService executor = Executors.newCachedThreadPool();

for (int i = 0; i < 10; i++) {
    final int threadNum = i;
    Thread.sleep(1000);
    executor.execute(() -> {
        try {
            race(threadNum);
        } catch (Exception e) {
            log.error("exception", e);
        }
    });
}
executor.shutdown();

}

private static void race(int threadNum) throws Exception {
Thread.sleep(1000);
log.info("{} is ready", threadNum);
barrier.await();
log.info("{} continue", threadNum);
}
}

![](https://upload-images.jianshu.io/upload_images/4685968-4fb51fa4926fd70e.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)

package com.mmall.concurrency.example.aqs;

import lombok.extern.slf4j.Slf4j;

import java.util.concurrent.CyclicBarrier;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;

/**

@author shishushengbr/>*/
@Slf4j
public class CyclicBarrierExample2 {

private static CyclicBarrier barrier = new CyclicBarrier(5);

public static void main(String[] args) throws Exception {

ExecutorService executor = Executors.newCachedThreadPool();

for (int i = 0; i < 10; i++) {
    final int threadNum = i;
    Thread.sleep(1000);
    executor.execute(() -> {
        try {
            race(threadNum);
        } catch (Exception e) {
            log.error("exception", e);
        }
    });
}
executor.shutdown();

}

private static void race(int threadNum) throws Exception {
Thread.sleep(1000);
log.info("{} is ready", threadNum);
try {
barrier.await(2000, TimeUnit.MILLISECONDS);
} catch (Exception e) {
log.warn("BarrierException", e);
}
log.info("{} continue", threadNum);
}
}

![await 超時導致程序拋異常](https://upload-images.jianshu.io/upload_images/4685968-0f899c23531f8ee8.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)

package com.mmall.concurrency.example.aqs;

import lombok.extern.slf4j.Slf4j;

import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.Semaphore;
/**

@author shishushengbr/>*/
@Slf4j
public class SemaphoreExample3 {

private final static int threadCount = 20;

public static void main(String[] args) throws Exception {

ExecutorService exec = Executors.newCachedThreadPool();

final Semaphore semaphore = new Semaphore(3);

for (int i = 0; i < threadCount; i++) {
    final int threadNum = i;
    exec.execute(() -> {
        try {
            // 嘗試獲取一個許可
            if (semaphore.tryAcquire()) {
                test(threadNum);
                // 釋放一個許可
                semaphore.release();
            }
        } catch (Exception e) {
            log.error("exception", e);
        }
    });
}
exec.shutdown();

}

private static void test(int threadNum) throws Exception {
log.info("{}", threadNum);
Thread.sleep(1000);
}

}


#9 線程池
##9.1 newCachedThreadPool
![](https://upload-images.jianshu.io/upload_images/4685968-1122da7a48223ba1.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
##9.2 newFixedThreadPool
![](https://upload-images.jianshu.io/upload_images/4685968-0ea942bf12e5210f.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
##9.3 newSingleThreadExecutor
看出是順序執行的
![](https://upload-images.jianshu.io/upload_images/4685968-989d59429f589403.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
##9.4 newScheduledThreadPool
![](https://upload-images.jianshu.io/upload_images/4685968-f7536ec7a1cf6ecc.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
![](https://upload-images.jianshu.io/upload_images/4685968-c90e09d5bfe707e6.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
#10 死鎖
![](https://upload-images.jianshu.io/upload_images/4685968-461f6a4251ae8ca4.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
![](https://upload-images.jianshu.io/upload_images/4685968-46d58773e597195f.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)

長文慎入-探索Java併發編程與高併發解決方案

Linux內核實戰（二）- 操作系統概覽

Linux內核實戰（一）- 學習路線規劃

macOS下 Hive 2.x 的安裝與配置

Spark機器學習實戰 (十二) - 推薦系統實戰

Spark機器學習實戰 (十一) - 文本情感分類項目實戰

https://yachay.unat.edu.pe/blog/index.php?comment_area=format_blog&comment_component=blog&comment_co

linux以太網驅動總結