7 种提升 Spring Boot 吞吐量

一、异步执行

实现方式二种：

1. 使用异步注解@aysnc、启动类：添加@EnableAsync注解2. JDK 8本身有一个非常好用的Future类——CompletableFuture

@AllArgsConstructor
public class AskThread implements Runnable{
    private CompletableFuture<Integer> re = null;

    public void run() {
        int myRe = 0;
        try {
            myRe = re.get() * re.get();
        } catch (Exception e) {
            e.printStackTrace();
        }
        System.out.println(myRe);
    }

    public static void main(String[] args) throws InterruptedException {
        final CompletableFuture<Integer> future = new CompletableFuture<>();
        new Thread(new AskThread(future)).start();
        //模拟长时间的计算过程
        Thread.sleep(1000);
        //告知完成结果
        future.complete(60);
    }
}

在该示例中，启动一个线程，此时AskThread对象还没有拿到它需要的数据，执行到 myRe = re.get() * re.get()会阻塞。我们用休眠1秒来模拟一个长时间的计算过程，并将计算结果告诉future执行结果，AskThread线程将会继续执行。如果您正在学习Spring Boot，那么推荐一个连载多年还在继续更新的免费教程：http://blog.didispace.com/spring-boot-learning-2x/

public class Calc {
    public static Integer calc(Integer para) {
        try {
            //模拟一个长时间的执行
            Thread.sleep(1000);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        return para * para;
    }

    public static void main(String[] args) throws ExecutionException, InterruptedException {
        final CompletableFuture<Void> future = CompletableFuture.supplyAsync(() -> calc(50))
                .thenApply((i) -> Integer.toString(i))
                .thenApply((str) -> "\"" + str + "\"")
                .thenAccept(System.out::println);
        future.get();
    }
}

CompletableFuture.supplyAsync方法构造一个CompletableFuture实例，在supplyAsync()方法中，它会在一个新线程中，执行传入的参数。在这里它会执行calc()方法，这个方法可能是比较慢的，但这并不影响CompletableFuture实例的构造速度，supplyAsync()会立即返回。

而返回的CompletableFuture实例就可以作为这次调用的契约，在将来任何场合，用于获得最终的计算结果supplyAsync用于提供返回值的情况，CompletableFuture还有一个不需要返回值的异步调用方法runAsync(Runnable runnable),一般我们在优化Controller时，使用这个方法比较多。这两个方法如果在不指定线程池的情况下，都是在ForkJoinPool.common线程池中执行，而这个线程池中的所有线程都是Daemon（守护）线程，所以，当主线程结束时，这些线程无论执行完毕都会退出系统。

核心代码：

CompletableFuture.runAsync(() ->
   this.afterBetProcessor(betRequest,betDetailResult,appUser,id)
);

异步调用使用Callable来实现

@RestController  
public class HelloController {
  
    private static final Logger logger = LoggerFactory.getLogger(HelloController.class);
      
    @Autowired  
    private HelloService hello;
  
    @GetMapping("/helloworld")
    public String helloWorldController() {
        return hello.sayHello();
    }
  
    /**
     * 异步调用restful
     * 当controller返回值是Callable的时候，springmvc就会启动一个线程将Callable交给TaskExecutor去处理
     * 然后DispatcherServlet还有所有的spring拦截器都退出主线程，然后把response保持打开的状态
     * 当Callable执行结束之后，springmvc就会重新启动分配一个request请求，然后DispatcherServlet就重新
     * 调用和处理Callable异步执行的返回结果， 然后返回视图
     *
     * @return
     */  
    @GetMapping("/hello")
    public Callable<String> helloController() {
        logger.info(Thread.currentThread().getName() + " 进入helloController方法");
        Callable<String> callable = new Callable<String>() {
  
            @Override  
            public String call() throws Exception {
                logger.info(Thread.currentThread().getName() + " 进入call方法");
                String say = hello.sayHello();
                logger.info(Thread.currentThread().getName() + " 从helloService方法返回");
                return say;
            }
        };
        logger.info(Thread.currentThread().getName() + " 从helloController方法返回");
        return callable;
    }
}

异步调用的方式 WebAsyncTask

@RestController  
public class HelloController {
  
    private static final Logger logger = LoggerFactory.getLogger(HelloController.class);
      
    @Autowired  
    private HelloService hello;
  
        /**
     * 带超时时间的异步请求 通过WebAsyncTask自定义客户端超时间
     *
     * @return
     */  
    @GetMapping("/world")
    public WebAsyncTask<String> worldController() {
        logger.info(Thread.currentThread().getName() + " 进入helloController方法");
  
        // 3s钟没返回，则认为超时
        WebAsyncTask<String> webAsyncTask = new WebAsyncTask<>(3000, new Callable<String>() {
  
            @Override  
            public String call() throws Exception {
                logger.info(Thread.currentThread().getName() + " 进入call方法");
                String say = hello.sayHello();
                logger.info(Thread.currentThread().getName() + " 从helloService方法返回");
                return say;
            }
        });
        logger.info(Thread.currentThread().getName() + " 从helloController方法返回");
  
        webAsyncTask.onCompletion(new Runnable() {
  
            @Override  
            public void run() {
                logger.info(Thread.currentThread().getName() + " 执行完毕");
            }
        });
  
        webAsyncTask.onTimeout(new Callable<String>() {
  
            @Override  
            public String call() throws Exception {
                logger.info(Thread.currentThread().getName() + " onTimeout");
                // 超时的时候，直接抛异常，让外层统一处理超时异常
                throw new TimeoutException("调用超时");
            }
        });
        return webAsyncTask;
    }
  
    /**
     * 异步调用，异常处理，详细的处理流程见MyExceptionHandler类
     *
     * @return
     */  
    @GetMapping("/exception")
    public WebAsyncTask<String> exceptionController() {
        logger.info(Thread.currentThread().getName() + " 进入helloController方法");
        Callable<String> callable = new Callable<String>() {
  
            @Override  
            public String call() throws Exception {
                logger.info(Thread.currentThread().getName() + " 进入call方法");
                throw new TimeoutException("调用超时!");
            }
        };
        logger.info(Thread.currentThread().getName() + " 从helloController方法返回");
        return new WebAsyncTask<>(20000, callable);
    }
  
}

二、增加内嵌Tomcat的最大连接数

@Configuration
public class TomcatConfig {
    @Bean
    public ConfigurableServletWebServerFactory webServerFactory() {
        TomcatServletWebServerFactory tomcatFactory = new TomcatServletWebServerFactory();
        tomcatFactory.addConnectorCustomizers(new MyTomcatConnectorCustomizer());
        tomcatFactory.setPort(8005);
        tomcatFactory.setContextPath("/api-g");
        return tomcatFactory;
    }
    class MyTomcatConnectorCustomizer implements TomcatConnectorCustomizer {
        public void customize(Connector connector) {
            Http11NioProtocol protocol = (Http11NioProtocol) connector.getProtocolHandler();
            //设置最大连接数
            protocol.setMaxConnections(20000);
            //设置最大线程数
            protocol.setMaxThreads(2000);
            protocol.setConnectionTimeout(30000);
        }
    }

}

三、使用@ComponentScan()定位扫包比@SpringBootApplication扫包更快

四、默认tomcat容器改为Undertow（Jboss下的服务器，Tomcat吞吐量5000，Undertow吞吐量8000）

<exclusions>
  <exclusion>
     <groupId>org.springframework.boot</groupId>
     <artifactId>spring-boot-starter-tomcat</artifactId>
  </exclusion>
</exclusions>

改为:

<dependency>
  <groupId>org.springframework.boot</groupId>
  <artifactId>spring-boot-starter-undertow</artifactId>
</dependency>

五、使用 BufferedWriter 进行缓冲

六、Deferred方式实现异步调用

@RestController
public class AsyncDeferredController {
    private final Logger logger = LoggerFactory.getLogger(this.getClass());
    private final LongTimeTask taskService;
    
    @Autowired
    public AsyncDeferredController(LongTimeTask taskService) {
        this.taskService = taskService;
    }
    
    @GetMapping("/deferred")
    public DeferredResult<String> executeSlowTask() {
        logger.info(Thread.currentThread().getName() + "进入executeSlowTask方法");
        DeferredResult<String> deferredResult = new DeferredResult<>();
        // 调用长时间执行任务
        taskService.execute(deferredResult);
        // 当长时间任务中使用deferred.setResult("world");这个方法时，会从长时间任务中返回，继续controller里面的流程
        logger.info(Thread.currentThread().getName() + "从executeSlowTask方法返回");
        // 超时的回调方法
        deferredResult.onTimeout(new Runnable(){
  
   @Override
   public void run() {
    logger.info(Thread.currentThread().getName() + " onTimeout");
    // 返回超时信息
    deferredResult.setErrorResult("time out!");
   }
  });
        
        // 处理完成的回调方法，无论是超时还是处理成功，都会进入这个回调方法
        deferredResult.onCompletion(new Runnable(){
  
   @Override
   public void run() {
    logger.info(Thread.currentThread().getName() + " onCompletion");
   }
  });
        
        return deferredResult;
    }
}

七、异步调用可以使用AsyncHandlerInterceptor进行拦截

@Component
public class MyAsyncHandlerInterceptor implements AsyncHandlerInterceptor {
 
 private static final Logger logger = LoggerFactory.getLogger(MyAsyncHandlerInterceptor.class);
 
 @Override
 public boolean preHandle(HttpServletRequest request, HttpServletResponse response, Object handler)
   throws Exception {
  return true;
 }
 
 @Override
 public void postHandle(HttpServletRequest request, HttpServletResponse response, Object handler,
   ModelAndView modelAndView) throws Exception {
// HandlerMethod handlerMethod = (HandlerMethod) handler;
  logger.info(Thread.currentThread().getName()+ "服务调用完成，返回结果给客户端");
 }
 
 @Override
 public void afterCompletion(HttpServletRequest request, HttpServletResponse response, Object handler, Exception ex)
   throws Exception {
  if(null != ex){
   System.out.println("发生异常:"+ex.getMessage());
  }
 }
 
 @Override
 public void afterConcurrentHandlingStarted(HttpServletRequest request, HttpServletResponse response, Object handler)
   throws Exception {
  
  // 拦截之后，重新写回数据，将原来的hello world换成如下字符串
  String resp = "my name is chhliu!";
  response.setContentLength(resp.length());
  response.getOutputStream().write(resp.getBytes());
  
  logger.info(Thread.currentThread().getName() + " 进入afterConcurrentHandlingStarted方法");
 }
 
}

参考

• https://my.oschina.net/u/3768341/blog/3001731

• https://blog.csdn.net/liuchuanhong1/article/details/78744138

来源：https://xhcom.blog.csdn.net/article/details/88046026