SpringBoot + Disruptor实现高并发内存消息队列

左手的ㄟ右手 2023-10-04 21:09 7阅读 0赞

1. 简介

Disruptor是英国外汇交易公司LMAX开发的一个高性能队列，研发的初衷是解决内存队列的延迟问题(在性能测试中发现竟然与I/O操作处于同样的数量级)。基于Disruptor开发的系统单线程能支撑每秒600万订单，2010年在QCon演讲后，获得了业界关注。  
  Disruptor区别于Kafka、RabbitMQ等消息队列，它是一个高性能的线程间异步通信的框架，即在同一个JVM进程中的多线程间消息传递。和ArrayBlockingQueue比较类似，但是它是一个有界无锁的高并发队列，如果项目中使用ArrayBlockingQueue在多线程之间传递消息，可以考虑是用Disruptor来代替。  
  Github：[GitHub - LMAX-Exchange/disruptor: High Performance Inter-Thread Messaging Library][GitHub - LMAX-Exchange_disruptor_ High Performance Inter-Thread Messaging Library]  
  文档：[LMAX Disruptor][]  
  资料及测试结论：[LMAX Disruptor: High performance alternative to bounded queues for exchanging data between concurrent threads][LMAX Disruptor_ High performance alternative to bounded queues for exchanging data between concurrent threads]

2. 官方测试

*  吞吐量性能测试

>   参考ArrayBlockingQueue(简称ABQ)在不同机器、不同消费模式(下文详细介绍)环境下的性能对比，详情浏览官方测试报告：[LMAX Disruptor: High performance alternative to bounded queues for exchanging data between concurrent threads][LMAX Disruptor_ High performance alternative to bounded queues for exchanging data between concurrent threads 1]

<table> 
 <thead> 
  <tr> 
   <th></th> 
   <th colspan="2">Nehalem 2.8Ghz – Windows 7 SP1 64-bit</th> 
   <th colspan="2">Sandy Bridge 2.2Ghz – Linux 2.6.38 64-bit</th> 
  </tr> 
  <tr> 
   <th></th> 
   <th>ABQ</th> 
   <th>Disruptor</th> 
   <th>ABQ</th> 
   <th>Disruptor</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>Unicast: 1P – 1C</td> 
   <td>5,339,256</td> 
   <td>25,998,336</td> 
   <td>4,057,453</td> 
   <td>22,381,378</td> 
  </tr> 
  <tr> 
   <td>Pipeline: 1P – 3C</td> 
   <td>2,128,918</td> 
   <td>16,806,157</td> 
   <td>2,006,903</td> 
   <td>15,857,913</td> 
  </tr> 
  <tr> 
   <td>Sequencer: 3P – 1C</td> 
   <td>5,539,531</td> 
   <td>13,403,268</td> 
   <td>2,056,118</td> 
   <td>14,540,519</td> 
  </tr> 
  <tr> 
   <td>Multicast: 1P – 3C</td> 
   <td>1,077,384</td> 
   <td>9,377,871</td> 
   <td>260,733</td> 
   <td>10,860,121</td> 
  </tr> 
  <tr> 
   <td>Diamond: 1P – 3C</td> 
   <td>2,113,941</td> 
   <td>16,143,613</td> 
   <td>2,082,725</td> 
   <td>15,295,197</td> 
  </tr> 
 </tbody> 
</table>

*  延迟性能测试

>   参考ArrayBlockingQueue(简称ABQ)在相同机器、相同同消费模式环境下的对比，详情浏览官方测试报告：[https://lmax-exchange.github.io/disruptor/disruptor.html\#\_latency\_performance\_testing][https_lmax-exchange.github.io_disruptor_disruptor.html_latency_performance_testing]

<table> 
 <thead> 
  <tr> 
   <th></th> 
   <th>Array Blocking Queue (ns)</th> 
   <th>Disruptor (ns)</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>Min Latency</td> 
   <td>145</td> 
   <td>29</td> 
  </tr> 
  <tr> 
   <td>Mean Latency</td> 
   <td>32,757</td> 
   <td>52</td> 
  </tr> 
  <tr> 
   <td>99% observations less than</td> 
   <td>2,097,152</td> 
   <td>128</td> 
  </tr> 
  <tr> 
   <td>99.99% observations less than</td> 
   <td>4,194,304</td> 
   <td>8,192</td> 
  </tr> 
  <tr> 
   <td>Max Latency</td> 
   <td>5,069,086</td> 
   <td>175,567</td> 
  </tr> 
 </tbody> 
</table>

3. 相关博客 [【并发编程】Java 阻塞队列![icon-default.png_t_N176][]https://www.cnblogs.com/cao-lei/p/14918871.html][Java _icon-default.png_t_N176_https_www.cnblogs.com_cao-lei_p_14918871.html][【并发编程】Java常用并发工具类![icon-default.png_t_N176][]https://www.cnblogs.com/cao-lei/p/14277053.html][Java_icon-default.png_t_N176_https_www.cnblogs.com_cao-lei_p_14277053.html][【并发编程】JAVA锁机制![icon-default.png_t_N176][]https://www.cnblogs.com/cao-lei/p/13219720.html][JAVA_icon-default.png_t_N176_https_www.cnblogs.com_cao-lei_p_13219720.html][【并发编程】JAVA CAS(Conmpare And Swap)原理![icon-default.png_t_N176][]https://www.cnblogs.com/cao-lei/p/13219106.html][JAVA CAS_Conmpare And Swap_icon-default.png_t_N176_https_www.cnblogs.com_cao-lei_p_13219106.html][【并发编程】Java并发编程：volatile关键字![icon-default.png_t_N176][]https://www.cnblogs.com/cao-lei/p/12937668.html][Java_volatile_icon-default.png_t_N176_https_www.cnblogs.com_cao-lei_p_12937668.html]

4. Java阻塞队列存在的弊端

常用的Java阻塞队列：

<table> 
 <thead> 
  <tr> 
   <th>名称</th> 
   <th>是否有界</th> 
   <th>是否加锁</th> 
   <th>数据结构</th> 
   <th>队列类型</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>ArrayBlockingQueue</td> 
   <td>有界</td> 
   <td>加锁</td> 
   <td>数组</td> 
   <td>阻塞</td> 
  </tr> 
  <tr> 
   <td>LinkedBlockingQueue</td> 
   <td>可选有界</td> 
   <td>加锁</td> 
   <td>链表</td> 
   <td>阻塞</td> 
  </tr> 
  <tr> 
   <td>PriorityBlockingQueue</td> 
   <td>无界</td> 
   <td>加锁</td> 
   <td>数组</td> 
   <td>阻塞</td> 
  </tr> 
  <tr> 
   <td>DelayQueue</td> 
   <td>无界</td> 
   <td>加锁</td> 
   <td>数组</td> 
   <td>阻塞</td> 
  </tr> 
  <tr> 
   <td>LinkedTransferQueue</td> 
   <td>无界</td> 
   <td>无锁</td> 
   <td>链表</td> 
   <td>阻塞</td> 
  </tr> 
  <tr> 
   <td>LinkedBlockingDeque</td> 
   <td>可选有界</td> 
   <td>有锁</td> 
   <td>链表</td> 
   <td>阻塞</td> 
  </tr> 
 </tbody> 
</table>

在使用中，为了防止生产过快而消费不及时导致的内存溢出，或垃圾回收频繁导致的性能问题，一般会使用有界且数据结构为数组的阻塞队列，即：ArrayBlockingQueue。但是，ArrayBlockingQueue使用加锁的方式保证线程安全，在低延迟的场景中表现悲观，且存在伪共享的问题，因此结果不尽人意。

5. 伪共享

5.1 CPU内部存储结构

现在的CPU都是多个CPU核心，如下图。为了提高访问效率，都有缓存行。每个核中都有L1 Cache和L2 Cache，L3 Cache则在多核之间共享。CPU在执行运算时，首先从L1 Cache查找数据，找不到则以一次从L2 Cache、L3 Cache查找，如果还是没有，则会去内存中查找，路径越长，耗时越长，性能越低。当数据被修改后，通过主线通知其他CPU将读取的数据标记为失效状态，下次访问时从内存重新读取数据到Cache。

![ebda0551f55149c1642b6934d4848860.jpeg][]  
  对于计算机的存储设备而言，除了CPU之外，外部还有内存和磁盘。如下图，存储容量越来越大，成本越来越大，但访问速度却越来越慢。![66be46274cbb5f6b3d03c181a23cb0bd.jpeg][]

5.2 缓存行

CPU从内存中加载数据时，并不是一个字节一个字节的加载，而是一块一块的的加载数据，这样的一块称为：缓存行(Cache Line)，即缓存行是CPU读取数据的最小单位。  
  CPU的缓存行一般为32 ~ 128字节，常见的CPU缓存行为64字节。

# 查询CPU的Cache Line大小
    cat /sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size

对于数组而言，CPU每次会加载数据中多个数据到Cache中。所以，如果按照物理内存地址分布的顺序去访问数据，Cache命中率最高，从而减少从内存加载数据的次数，提高性能。  
  但对于单个变量而言，会存在Cache伪共享的问题。假设定义衣蛾Long类型的变量A，占用8个字节，则CPU每次从内存中读取数剧是，会连同连续地址内的其余7个数据一并加载到Cache中。

![8003cbcd0cadf2d97979983a0e3296f6.jpeg][]

5.3 伪共享

多个CPU缓存遵循`MESI协议`。  
  假设，定义两个变量A、B，线程1绑定Core1，读取变量A，线程2绑定Core2，读取变量B。

![7b248fde00b7c65b6676628203fdf821.jpeg][]  
  Core1和Core2分别读取变量A、B到Cache中，但变量A、B在同一Cache Line，因此，Core1和Core2都会把数据A、B加载到Cache中。![9313eb7b9db0bbabd013304d94b0f36f.jpeg][]  
  线程1通过Core1修改变量A。首先通过主线发送消息给Core2将存放变量A的Cache Line标记为失效状态，然后将Core1中的Cache Line的状态标为已修改，并修改数据。![11cfc87d5b5216137fe24ab37ff8be52.jpeg][]  
  线程2修改变量B时，发现Cache Line的状态为失效，并且Core1中的Cache Line为已修改状态，则先把Core1中的变量A、B写回内存，然后从内存中重新读取变量A、B。然后通过主线发送消息给Core1将存放变量B的Cache Line标记为失效状态，然后将Core2中的Cache Line的状态标为已修改，并修改数据。

![d141b63c9e947de619e67adbe6049288.jpeg][]  
  如果线程1、2频繁交替修改变量A、B，则会重复以上步骤，导致Cache没有意义。虽然变量A、B之间没有任何关系，但属于同一Cache Line。  
  这种多个线程同时读写同一个Cache Line的不同变量而导致CPU Cache失效的现象称为伪共享(False Sharing)。

5.4 ArrayBlockingQueue中的伪共享

查看`ArrayBlockingQueue`源码，有三个核心变量：

# 出队下标
    /** items index for next take, poll, peek or remove */
    int takeIndex;
    
    # 入队下标
    /** items index for next put, offer, or add */
    int putIndex;
    
    # 队列中元素数量
    /** Number of elements in the queue */
    int count;

这三个变量很容易放到同一Cache Line，这也是影响`ArrayBlockingQueue`性能的重要原因之一。

6. Disruptor如何避免伪共享

`ArrayBlockingQueue`因为加锁和伪共享的问题影响性能，那`Disruptor`是如何避免这两个问题来提供性能的呢？主要表现在以下几个方面：

*  采用环形数组结构
 *  采用CAS无锁方式
 *  添加额外的信息避免伪共享

6.1 环形数组结构

环形数组(`RingBuffer`)结构是`Disruptor`的核心。官网对环形数据结构的剖析：[Dissecting the Disruptor: What’s so special about a ring buffer?][Dissecting the Disruptor_ What_s so special about a ring buffer]![f75a0e8f9bdcfc0aafed29b62ff96908.jpeg][]  
优势：

*  数组结构：当CPU Cache加载数据时，相邻的数据也会被加载到Cache中，避免CPU频繁从内存中获取数据。
 *  避免GC：实质还是一个普通的数组，当数据填满队列时(2^n - 1)时，再次添加数据回覆盖之前的数据。
 *  位运算：数组大小必须为2的n次方，通过位运算提高效率。

6.2 CAS无锁方式

`Disruptor`与传统的队列不同，分为队首指针和队尾指针，而是只有一个游标器`Sequencer`，它可以保证生产的消息不回覆盖还未消费的消息。`Sequencer`分为两个实现类：`SingleProducerSequencer`和`MultiProducerSequencer`，即单生产者和多生产者。当单个生产者时，生产者每次从`RingBuffer`中获取下一个可以生产的位置，然后存放数据；消费者先获取最大的可消费的位置，再读取数据进行消费。当多个生产者时，每个生产者下通过CAS竞争获取可以生产的位置，然后存放数据；每个消费者都需要先获取最大可消费的下标，然后读取数据进行消费。

6.3 添加额外信息

`RingBuffer`的下标是一个volatile变量，即不用加锁就能保证多线程安全，同时每个long类型的下标还会附带7个long的额外变量，避免伪共享的问题。

class LhsPadding
    {
        protected long p1, p2, p3, p4, p5, p6, p7;
    }
    
    class Value extends LhsPadding
    {
        protected volatile long value;
    }
    
    class RhsPadding extends Value
    {
        protected long p9, p10, p11, p12, p13, p14, p15;
    }

7. 示例代码

*  创建项目![a942e2966acf9fdb59b9044cf69c82c6.jpeg][]
 *  修改pom文件

<?xml version="1.0" encoding="UTF-8"?>
    <project xmlns="http://maven.apache.org/POM/4.0.0"
             xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
             xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
        <modelVersion>4.0.0</modelVersion>
    
        <groupId>com.c3stones</groupId>
        <artifactId>springboot-disruptor-demo</artifactId>
        <version>1.0-SNAPSHOT</version>
    
        <dependencies>
            <dependency>
                <groupId>com.lmax</groupId>
                <artifactId>disruptor</artifactId>
                <version>3.4.4</version>
            </dependency>
            <dependency>
                <groupId>org.projectlombok</groupId>
                <artifactId>lombok</artifactId>
                <version>1.18.24</version>
                <optional>true</optional>
            </dependency>
            <dependency>
                <groupId>org.slf4j</groupId>
                <artifactId>slf4j-api</artifactId>
                <version>1.7.36</version>
            </dependency>
            <dependency>
                <groupId>org.slf4j</groupId>
                <artifactId>slf4j-simple</artifactId>
                <version>1.7.36</version>
            </dependency>
            <dependency>
                <groupId>junit</groupId>
                <artifactId>junit</artifactId>
                <version>4.12</version>
                <scope>test</scope>
            </dependency>
        </dependencies>
    
        <build>
            <plugins>
                <plugin>
                    <groupId>org.apache.maven.plugins</groupId>
                    <artifactId>maven-compiler-plugin</artifactId>
                    <configuration>
                        <source>8</source>
                        <target>8</target>
                    </configuration>
                </plugin>
            </plugins>
        </build>
    
    </project>

*  创建消息模型类  
      用于在`Disruptor`之间传递消息。

import lombok.AllArgsConstructor;
    import lombok.Data;
    import lombok.NoArgsConstructor;
    
    /**
     * 消息
     *
     * @author CL
     */
    @Data
    @NoArgsConstructor
    @AllArgsConstructor
    public class Message {
    
        private Object data;
    
        @Override
        public String toString() {
            return data.toString();
        }
    
    }

*  创建生产者

import com.c3stones.model.Message;
    import com.lmax.disruptor.RingBuffer;
    import com.lmax.disruptor.dsl.Disruptor;
    import lombok.RequiredArgsConstructor;
    
    /**
     * 生产者
     *
     * @author CL
     */
    @RequiredArgsConstructor
    public class Producer {
    
        private final Disruptor disruptor;
    
        /**
         * 发送数据
         *
         * @param data 数据
         */
        public void send(Object data) {
            RingBuffer<Message> ringBuffer = disruptor.getRingBuffer();
            // 获取可以生成的位置
            long next = ringBuffer.next();
            try {
                Message msg = ringBuffer.get(next);
                msg.setData(data);
            } finally {
                ringBuffer.publish(next);
            }
        }
    
    }

*  创建分组模式消费者

import com.c3stones.model.Message;
    import com.lmax.disruptor.WorkHandler;
    import lombok.RequiredArgsConstructor;
    import lombok.extern.slf4j.Slf4j;
    
    /**
     * 分组消费
     *
     * @author CL
     */
    @Slf4j
    @RequiredArgsConstructor
    public class GroupConsumer implements WorkHandler<Message> {
    
        /**
         * 消费着编号
         */
        private final Integer number;
    
        /**
         * 分组消费：每个生产者生产的数据只能被一个消费者消费
         *
         * @param message 消息
         */
        @Override
        public void onEvent(Message message) {
            log.info("Group Consumer number: {}, message: {}", number, message);
        }
    
    }

*  创建重复模式消费者

import com.c3stones.model.Message;
    import com.lmax.disruptor.EventHandler;
    import lombok.RequiredArgsConstructor;
    import lombok.extern.slf4j.Slf4j;
    
    /**
     * 重复消费
     *
     * @author CL
     */
    @Slf4j
    @RequiredArgsConstructor
    public class RepeatConsumer implements EventHandler<Message> {
    
        /**
         * 消费着编号
         */
        private final Integer number;
    
        /**
         * 重复消费：每个消费者重复消费生产者生产的数据
         *
         * @param message    消息
         * @param sequence   当前序列号
         * @param endOfBatch 批次结束标识(常用于将多个消费着的数据依次组合到最后一个消费者统一处理)
         */
        @Override
        public void onEvent(Message message, long sequence, boolean endOfBatch) {
            log.info("Repeat Consumer number: {}, message: {}, curr sequence: {}, is end: {}",
                    number, message, sequence, endOfBatch);
        }
    
    }

*  创建通用模式消费者  
      一般采用这种更加通用的模式。

import com.c3stones.model.Message;
    import com.lmax.disruptor.EventHandler;
    import com.lmax.disruptor.WorkHandler;
    import lombok.RequiredArgsConstructor;
    import lombok.extern.slf4j.Slf4j;
    
    /**
     * 消费者
     *
     * @author CL
     */
    @Slf4j
    @RequiredArgsConstructor
    public class Consumer implements WorkHandler<Message>, EventHandler<Message> {
    
        private final String desc;
    
        /**
         * 重复消费：每个消费者重复消费生产者生产的数据
         *
         * @param message    消息
         * @param sequence   当前序列号
         * @param endOfBatch 批次结束标识(常用于将多个消费着的数据依次组合到最后一个消费者统一处理)
         */
        @Override
        public void onEvent(Message message, long sequence, boolean endOfBatch) {
            this.onEvent(message);
        }
    
        /**
         * 分组消费：每个生产者生产的数据只能被一个消费者消费
         *
         * @param message 消息
         */
        @Override
        public void onEvent(Message message) {
            log.info("Group Consumer describe: {}, message: {}", desc, message);
        }
    
    }

8. 单元测试

*  测试分组消费：每个生产者生产的数据只能被一个消费者消费

import com.c3stones.consumer.GroupConsumer;
    import com.c3stones.model.Message;
    import com.c3stones.producer.Producer;
    import com.lmax.disruptor.dsl.Disruptor;
    import org.junit.After;
    import org.junit.Before;
    import org.junit.Test;
    
    import java.util.Arrays;
    import java.util.concurrent.Executors;
    
    /**
     * 分组消费 单元测试
     *
     * @author CL
     */
    public class GroupConsumerTest {
    
        private Disruptor<Message> disruptor;
    
        @Before
        public void init() {
            GroupConsumer a = new GroupConsumer(1);
            GroupConsumer b = new GroupConsumer(2);
    
            disruptor = new Disruptor(Message::new, 1024, Executors.defaultThreadFactory());
    
            disruptor.handleEventsWithWorkerPool(a, b);
    
            disruptor.start();
        }
    
        @After
        public void close() {
            disruptor.shutdown();
        }
    
        @Test
        public void test() {
            Producer producer = new Producer(disruptor);
    
            Arrays.asList("aaa", "bbb").forEach(data -> producer.send(data));
        }
    
    }

单元测试结果：

[pool-1-thread-1] INFO com.c3stones.consumer.GroupConsumer - Group Consumer number: 1, message: bbb
    [pool-1-thread-2] INFO com.c3stones.consumer.GroupConsumer - Group Consumer number: 2, message: aaa

*  测试重复消费：每个消费者重复消费生产者生产的数据

import com.c3stones.consumer.RepeatConsumer;
    import com.c3stones.model.Message;
    import com.c3stones.producer.Producer;
    import com.lmax.disruptor.dsl.Disruptor;
    import org.junit.After;
    import org.junit.Before;
    import org.junit.Test;
    
    import java.util.Arrays;
    import java.util.concurrent.Executors;
    
    /**
     * 重复消费 单元测试
     *
     * @author CL
     */
    public class RepeatConsumerTest {
    
        private Disruptor<Message> disruptor;
    
        @Before
        public void init() {
            RepeatConsumer a = new RepeatConsumer(1);
            RepeatConsumer b = new RepeatConsumer(2);
    
            disruptor = new Disruptor(Message::new, 1024, Executors.defaultThreadFactory());
    
            disruptor.handleEventsWith(a, b);
    
            disruptor.start();
        }
    
        @After
        public void close() {
            disruptor.shutdown();
        }
    
        @Test
        public void test() {
            Producer producer = new Producer(disruptor);
    
            Arrays.asList("aaa", "bbb").forEach(data -> producer.send(data));
        }
    
    }

单元测试结果：

[pool-1-thread-1] INFO com.c3stones.consumer.RepeatConsumer - Repeat Consumer number: 1, message: aaa, curr sequence: 0, is end: false
    [pool-1-thread-2] INFO com.c3stones.consumer.RepeatConsumer - Repeat Consumer number: 2, message: aaa, curr sequence: 0, is end: false
    [pool-1-thread-1] INFO com.c3stones.consumer.RepeatConsumer - Repeat Consumer number: 1, message: bbb, curr sequence: 1, is end: true
    [pool-1-thread-2] INFO com.c3stones.consumer.RepeatConsumer - Repeat Consumer number: 2, message: bbb, curr sequence: 1, is end: true

*  测试链路模式

/**
     * 测试链路模式
     * <p>
     * p => a -> b -> c
     * </p>
     */
    @Test
    public void testChain() throws InterruptedException {
        Consumer a = new Consumer("a");
        Consumer b = new Consumer("b");
        Consumer c = new Consumer("c");
    
        Disruptor<Message> disruptor = new Disruptor(Message::new, 1024, Executors.defaultThreadFactory());
    
        disruptor.handleEventsWith(a).then(b).then(c);
    
        disruptor.start();
    
        Producer producer = new Producer(disruptor);
    
        producer.send("Chain");
    
        Thread.sleep(1000);
    
        disruptor.shutdown();
    }

单元测试结果：

[pool-1-thread-1] INFO com.c3stones.consumer.Consumer - Group Consumer describe: a, message: Chain
    [pool-1-thread-2] INFO com.c3stones.consumer.Consumer - Group Consumer describe: b, message: Chain
    [pool-1-thread-3] INFO com.c3stones.consumer.Consumer - Group Consumer describe: c, message: Chain

*  测试钻石模式

/**
         * 测试钻石模式
         * <p>
         *           a
         *        ↗     ↘
         *  p =>            c
         *        ↘     ↗
         *           b
         * </p>
         */
        @Test
        public void testDiamond() throws InterruptedException {
            Consumer a = new Consumer("a");
            Consumer b = new Consumer("b");
            Consumer c = new Consumer("c");
    
            Disruptor<Message> disruptor = new Disruptor(Message::new, 1024, Executors.defaultThreadFactory());
    
            disruptor.handleEventsWithWorkerPool(a, b).then(c);
    
            disruptor.start();
    
            Producer producer = new Producer(disruptor);
    
            producer.send("Diamond1");
            producer.send("Diamond2");
    
            Thread.sleep(1000);
    
            disruptor.shutdown();
        }

单元测试结果：

[pool-1-thread-1] INFO com.c3stones.consumer.Consumer - Group Consumer describe: a, message: Diamond2
    [pool-1-thread-2] INFO com.c3stones.consumer.Consumer - Group Consumer describe: b, message: Diamond1
    [pool-1-thread-3] INFO com.c3stones.consumer.Consumer - Group Consumer describe: c, message: Diamond1
    [pool-1-thread-3] INFO com.c3stones.consumer.Consumer - Group Consumer describe: c, message: Diamond2

9. 项目地址

[spring-boot-disuptor-demo][]

[GitHub - LMAX-Exchange_disruptor_ High Performance Inter-Thread Messaging Library]: https://github.com/LMAX-Exchange/disruptor
[LMAX Disruptor]: https://lmax-exchange.github.io/disruptor/
[LMAX Disruptor_ High performance alternative to bounded queues for exchanging data between concurrent threads]: https://lmax-exchange.github.io/disruptor/disruptor.html#_overview
[LMAX Disruptor_ High performance alternative to bounded queues for exchanging data between concurrent threads 1]: https://lmax-exchange.github.io/disruptor/disruptor.html#_throughput_performance_testing
[https_lmax-exchange.github.io_disruptor_disruptor.html_latency_performance_testing]: https://lmax-exchange.github.io/disruptor/disruptor.html#_latency_performance_testing
[icon-default.png_t_N176]: https://csdnimg.cn/release/blog_editor_html/release2.2.3/ckeditor/plugins/CsdnLink/icons/icon-default.png?t=N176
[Java _icon-default.png_t_N176_https_www.cnblogs.com_cao-lei_p_14918871.html]: https://www.cnblogs.com/cao-lei/p/14918871.html
[Java_icon-default.png_t_N176_https_www.cnblogs.com_cao-lei_p_14277053.html]: https://www.cnblogs.com/cao-lei/p/14277053.html
[JAVA_icon-default.png_t_N176_https_www.cnblogs.com_cao-lei_p_13219720.html]: https://www.cnblogs.com/cao-lei/p/13219720.html
[JAVA CAS_Conmpare And Swap_icon-default.png_t_N176_https_www.cnblogs.com_cao-lei_p_13219106.html]: https://www.cnblogs.com/cao-lei/p/13219106.html
[Java_volatile_icon-default.png_t_N176_https_www.cnblogs.com_cao-lei_p_12937668.html]: https://www.cnblogs.com/cao-lei/p/12937668.html
[ebda0551f55149c1642b6934d4848860.jpeg]: https://img-blog.csdnimg.cn/img_convert/ebda0551f55149c1642b6934d4848860.jpeg
[66be46274cbb5f6b3d03c181a23cb0bd.jpeg]: https://img-blog.csdnimg.cn/img_convert/66be46274cbb5f6b3d03c181a23cb0bd.jpeg
[8003cbcd0cadf2d97979983a0e3296f6.jpeg]: https://img-blog.csdnimg.cn/img_convert/8003cbcd0cadf2d97979983a0e3296f6.jpeg
[7b248fde00b7c65b6676628203fdf821.jpeg]: https://img-blog.csdnimg.cn/img_convert/7b248fde00b7c65b6676628203fdf821.jpeg
[9313eb7b9db0bbabd013304d94b0f36f.jpeg]: https://img-blog.csdnimg.cn/img_convert/9313eb7b9db0bbabd013304d94b0f36f.jpeg
[11cfc87d5b5216137fe24ab37ff8be52.jpeg]: https://img-blog.csdnimg.cn/img_convert/11cfc87d5b5216137fe24ab37ff8be52.jpeg
[d141b63c9e947de619e67adbe6049288.jpeg]: https://img-blog.csdnimg.cn/img_convert/d141b63c9e947de619e67adbe6049288.jpeg
[Dissecting the Disruptor_ What_s so special about a ring buffer]: https://trishagee.com/2011/06/22/dissecting_the_disruptor_whats_so_special_about_a_ring_buffer/
[f75a0e8f9bdcfc0aafed29b62ff96908.jpeg]: https://img-blog.csdnimg.cn/img_convert/f75a0e8f9bdcfc0aafed29b62ff96908.jpeg
[a942e2966acf9fdb59b9044cf69c82c6.jpeg]: https://img-blog.csdnimg.cn/img_convert/a942e2966acf9fdb59b9044cf69c82c6.jpeg
[spring-boot-disuptor-demo]: https://github.com/C3Stones/blog