使用多线程提高rest服务性能
tomcat管理线程数量有限,当达到一定请求数量时,无法继续接受请求,使用多线程的方式,可以调用一个异步线程来执行。
执行逻辑如下图,tomcat就收http请求,调用一个副线程进行处理,副线程处理后,将结果返回给主线程。在副线程处理整个业务逻辑的过程中,主线程可以空闲出来,去处理其他请求。使得服务器的吞吐量可以有一个很大的提升。
用同步方式和异步方式编写两个请求(github地址)
package com.ustc.reed.controller.async;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.web.bind.annotation.GetMapping;
import org.springframework.web.bind.annotation.RestController;
/**
* @author reed fan
*
*/
import java.util.concurrent.Callable;
@RestController
public class AsyncController {
private static Logger logger = LoggerFactory.getLogger(AsyncController.class);
@GetMapping("/sync")
public String sync() throws Exception{
logger.info("主线程开始");
Thread.sleep(1000);
logger.info("主线程结束");
return "success";
}
@GetMapping("/async")
public Callable<String> async() throws Exception{
logger.info("主线程开始");
Callable<String> result = new Callable<String>() {
@Override
public String call() throws Exception {
logger.info("副线程开始");
Thread.sleep(1000);
logger.info("副线程返回");
return "success";
}
};
logger.info("主线程结束");
return result;
}
}
在浏览器中输入http://localhost:8082/reed/sync
在浏览器中输入http://localhost:8082/reed/async
---
在实际开发中,可能遇到接收请求和响应请求不是同一个线程的场景。如下图,此时使用callable无法满足业务需求。
可以使用 DeferredResult来处理此类业务场景。
以下单处理场景为例,应用服务器1的线程1收到下单请求,将下单信息发送给消息队列。应用服务器2消费消息,进行下单处理。下单完成后,将结果返回给消息队列, 应用服务器有另外一个线程2来监听消息队列,当发现有订单处理结果的消息,根据消息的结果返回HTTP响应。
线程1和线程2是完全隔离的,谁也不知道对方的存在。
-----
限于篇幅,新建一个MockQueue,来模拟下单处理。
package com.ustc.reed.service.sync;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.stereotype.Component;
/**
* created by reedfan on 2019/1/6
*/
@Component
public class MockQueue {
private String placeOrder; //下单消息
private String completeOrder; //下单成功消息
private Logger logger = LoggerFactory.getLogger(getClass());
public String getPlaceOrder() {
return placeOrder;
}
public void setPlaceOrder(String placeOrder) throws Exception {
new Thread(() -> {
logger.info("接到下单请求, " + placeOrder);
try {
Thread.sleep(1000);
} catch (Exception e) {
e.printStackTrace();
}
this.completeOrder = placeOrder;
logger.info("下单请求处理完毕," + placeOrder);
}).start();
}
public String getCompleteOrder() {
return completeOrder;
}
public void setCompleteOrder(String completeOrder) {
this.completeOrder = completeOrder;
}
}
------
每个订单号会有一个处理结果。DeferredResultHolder可以在图示的线程1和线程2之间传递DeferredResult这个对象map的key可以理解为订单号。
package com.ustc.reed.service.sync;
import org.springframework.stereotype.Component;
import org.springframework.web.context.request.async.DeferredResult;
import java.util.HashMap;
import java.util.Map;
/**
* created by reedfan on 2019/1/6
*/
@Component
public class DeferredResultHolder {
private Map<String,DeferredResult<String>> map = new HashMap<>();
public Map<String, DeferredResult<String>> getMap() {
return map;
}
public void setMap(Map<String, DeferredResult<String>> map) {
this.map = map;
}
}
-------
监听下单是否完成,完成则返回订单结果。
package com.ustc.reed.service.sync;
import org.apache.commons.lang3.StringUtils;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.context.ApplicationListener;
import org.springframework.context.event.ContextRefreshedEvent;
import org.springframework.stereotype.Component;
/**
* created by reedfan on 2019/1/6
*/
@Component
public class QueueListener implements ApplicationListener<ContextRefreshedEvent> {
private Logger logger = LoggerFactory.getLogger(QueueListener.class);
@Autowired
private MockQueue mockQueue;
@Autowired
private DeferredResultHolder deferredResultHolder;
@Override
public void onApplicationEvent(ContextRefreshedEvent event) {
new Thread(
new Runnable() {
@Override
public void run() {
while (true){
if(StringUtils.isNotBlank(mockQueue.getCompleteOrder())){
String orderNumber = mockQueue.getCompleteOrder();
logger.info("返回订单结果:"+orderNumber);
deferredResultHolder.getMap().get(orderNumber).setResult("place order success");
mockQueue.setCompleteOrder(null);
}else {
try {
Thread.sleep(100);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}
}
}
).start();
}
}
请求代码如下。
package com.ustc.reed.controller.async;
import com.ustc.reed.service.sync.DeferredResultHolder;
import com.ustc.reed.service.sync.MockQueue;
import org.apache.commons.lang3.RandomStringUtils;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.web.bind.annotation.GetMapping;
import org.springframework.web.bind.annotation.RestController;
import org.springframework.web.context.request.async.DeferredResult;
/**
* @author reed fan
*
*/
import java.util.concurrent.Callable;
@RestController
public class AsyncController {
private static Logger logger = LoggerFactory.getLogger(AsyncController.class);
@Autowired
private MockQueue mockQueue;
@Autowired
private DeferredResultHolder deferredResultHolder;
@GetMapping("/mqasync")
public DeferredResult<String> mqasync() throws Exception{
logger.info("主线程开始");
String orderNumber = RandomStringUtils.randomNumeric(8);
mockQueue.setPlaceOrder(orderNumber);
DeferredResult<String> result = new DeferredResult<>();
deferredResultHolder.getMap().put(orderNumber, result);
return result;
}
}
在浏览器中输入http://localhost:8082/reed/mqasync ,模拟出编号为05804777的订单的处理过程,结果如下。