一、背景
项目中肯定会遇到异步调用其他方法的场景,比如有个计算过程,需要计算很多个指标的值,但是每个指标计算的效率快慢不同,如果采用同步执行的方式,运行这一个过程的时间是计算所有指标的时间之和。比如:
方法A:计算指标x,指标y,指标z的值,其中计算指标x需要1s,计算指标y需要2s,指标z需要3s。最终执行完方法A就是5s。
现在用异步的方式优化一下
方法A异步调用方法B,方法C,方法D,方法B,方法C,方法D分别计算指标x,指标y,指标z的值,那么最终执行完方法A的时间则是3s。
还有一种用途是当一个业务里面需要多个请求时,这时候异步并发请求所得到的回报远远是物有所值的。因为他是异步执行的,话不多说,一下是在springBoot里面使用并发请求;
二、spring boot中异步并发使用
2.1、appllication.yml
#****************集成Async线程池开始******************* async: # Async线程池 配置 executor: corepoolsize: 20 maxpoolsize: 25 queuecapacity: 40 keepaliveseconds: 200 threadnameprefix: appasync awaitterminationseconds: 60 #*****************集成Async线程池结束******************
2.2、配置线程池
@Configuration @EnableAsync public class ExecutorConfig { @Value("${async.executor.corepoolsize}") private Integer corePoolSize; @Value("${async.executor.maxpoolsize}") private Integer maxPoolSize; @Value("${async.executor.queuecapacity}") private Integer queueCapacity; @Value("${async.executor.keepaliveseconds}") private Integer keepAliveSeconds; @Value("${async.executor.threadnameprefix}") private String threadNamePrefix; @Value("${async.executor.awaitterminationseconds}") private Integer awaitTerminationSeconds; /** * 线程池 * * @return */ @Bean(name = "asyncExecutor") public Executor asyncExecutor() { ThreadPoolTaskExecutor executor = new ThreadPoolTaskExecutor(); // 基础线程数 corePoolSize: 10 executor.setCorePoolSize(corePoolSize); // 最大线程数 maxPoolSize: 15 executor.setMaxPoolSize(maxPoolSize); // 队列长度 queueCapacity: 25 executor.setQueueCapacity(queueCapacity); // 线程池维护线程所允许的空闲时间,单位为秒 keepAliveSeconds: 200 executor.setKeepAliveSeconds(keepAliveSeconds); // 线程名字 threadNamePrefix: appasync executor.setThreadNamePrefix(threadNamePrefix); executor.setRejectedExecutionHandler(new ThreadPoolExecutor.CallerRunsPolicy()); // 等待所有任务都完成再继续销毁其他的Bean executor.setWaitForTasksToCompleteOnShutdown(true); // 线程池中任务的等待时间,如果超过这个时候还没有销毁就强制销毁,以确保应用最后能够被关闭,而不是阻塞住 executor.setAwaitTerminationSeconds(awaitTerminationSeconds); executor.initialize(); return executor; } }
2.3、线程池监控(这个可有可无,主要是为了对线程池参数及时的调优)
@RestController @Slf4j @RequestMapping("/pubapi/asyncExecutor") public class AsyncExecutorController extends BaseController { @Resource(name = "asyncExecutor") private Executor asyncExecutor; @PostMapping("/monitor")public ResultBean<Map<String, Object>> getAsyncExecutorData() { ResultBean<Map<String, Object>> resultBean = ResultBeanUtil.error500(); if (asyncExecutor == null) { return resultBean; } try { ThreadPoolTaskExecutor executorTask = (ThreadPoolTaskExecutor) asyncExecutor; ThreadPoolExecutor executor = executorTask.getThreadPoolExecutor(); // 当前排队线程数 int queueSize = executor.getQueue().size(); // 当前活动线程数 int activeCount = executor.getActiveCount(); // 执行完线程数 long completedThreadCount = executor.getCompletedTaskCount(); // 总线程数 long taskCount = executor.getTaskCount(); // 初始线程数 int poolSize = executor.getPoolSize(); // 核心线程数 int corePoolSize = executor.getCorePoolSize(); // 线程池是否终止 boolean isTerminated = executor.isTerminated(); // 线城池是否关闭 boolean isShutdown = executor.isShutdown(); // 线程空闲时间 long keepAliveTime = executor.getKeepAliveTime(TimeUnit.MILLISECONDS); // 最大允许线程数 long maximumPoolSize = executor.getMaximumPoolSize(); // 线程池中存在的最大线程数 long largestPoolSize = executor.getLargestPoolSize(); Map<String, Object> threadPoolData = new HashMap<>(18); threadPoolData.put("当前排队线程数", queueSize); threadPoolData.put("当前活动线程数", activeCount); threadPoolData.put("执行完线程数", completedThreadCount); threadPoolData.put("总线程数", taskCount); threadPoolData.put("初始线程数", poolSize); threadPoolData.put("核心线程数", corePoolSize); threadPoolData.put("线程池是否终止", isTerminated); threadPoolData.put("线城池是否关闭", isShutdown); threadPoolData.put("线程空闲时间", keepAliveTime); threadPoolData.put("最大允许线程数", maximumPoolSize); threadPoolData.put("线程池中存在的最大线程数", largestPoolSize); InetAddress inetAddress = IdWorker.getLocalHostLANAddress(); Map<String, Object> resultData = new HashMap<>(4); resultData.put("ip", inetAddress.getHostAddress()); resultData.put("threadPoolData", threadPoolData); resultBean = ResultBeanUtil.success("请求成功!", resultData); } catch (Exception e) { e.printStackTrace(); } return resultBean; } }
2.4、代码中使用
public void getMap(){ /** * 先将耗时的、相互之间无依赖的操作先执行,由于其执行结果暂时不是特别关注,所以 */ Future<String> futureA = functionA(); Future<String> futureB = functionB(); /** * 执行其他的操作,其实functionA(),functionB()也在工作 */ aaa(); /** * 获取异步的结果,然后计算 */ try { String resultA =futureA.get(); String resuleB = futureB.get(); } catch (InterruptedException e) { e.printStackTrace(); } catch (ExecutionException e) { e.printStackTrace(); } } public Future<String> functionA (){ Future<String> future = null; try { Thread.sleep(5000); future = new AsyncResult<String>("functionA"); } catch (InterruptedException e) { e.printStackTrace(); } return future; } public Future<String> functionB (){ Future<String> future = null; try { Thread.sleep(3000); future = new AsyncResult<String>("functionB"); } catch (InterruptedException e) { e.printStackTrace(); } return future; } public void aaa(){ System.out.println("我是"); }
三、使用误区
在使用时候,
四、线程池选型