[java] Java 8에서 블로킹 백그라운드 로더를 작성하는 방법은 무엇입니까?

질문

Java 8에서 적절한 백그라운드 로더를 어떻게 작성합니까? 조항들:

  • 백그라운드에서 데이터를로드해야합니다
  • 로드 후 데이터가 표시되어야합니다
  • 데이터가로드되는 동안 더 이상 요청을 수락하지 않아야합니다.
  • 데이터가로드되는 동안 요청이있는 경우 특정 시간 초과 (예 : 5 초) 후에 다른로드를 예약해야합니다.

예를 들어 다시로드 요청을 수락하지만 요청으로 데이터베이스가 넘치지 않도록하는 것입니다.

MCVE

여기 MCVE가 있습니다. Thread.sleep을 2 초간 호출하여 로딩을 시뮬레이션하는 백그라운드 작업으로 구성됩니다. 작업은 매초마다 예약되므로 자연스럽게 백그라운드 로딩 작업이 겹치게되므로 피해야합니다.

public class LoadInBackgroundExample {

  /**
   * A simple background task which should perform the data loading operation. In this minimal example it simply invokes Thread.sleep
   */
  public static class BackgroundTask implements Runnable {

    private int id;

    public BackgroundTask(int id) {
      this.id = id;
    }

    /**
     * Sleep for a given amount of time to simulate loading.
     */
    @Override
    public void run() {

      try {

        System.out.println("Start #" + id + ": " + Thread.currentThread());

        long sleepTime = 2000;
        Thread.sleep( sleepTime);

      } catch (InterruptedException e) {
        e.printStackTrace();
      } finally {
        System.out.println("Finish #" + id + ": " + Thread.currentThread());
      }

    }
  }

  /**
   * CompletableFuture which simulates loading and showing data.
   * @param taskId Identifier of the current task
   */
  public static void loadInBackground( int taskId) {

    // create the loading task
    BackgroundTask backgroundTask = new BackgroundTask( taskId);

    // "load" the data asynchronously
    CompletableFuture<String> completableFuture = CompletableFuture.supplyAsync(new Supplier<String>() {

      @Override
      public String get() {

        CompletableFuture<Void> future = CompletableFuture.runAsync(backgroundTask);

        try {

          future.get();

        } catch (InterruptedException | ExecutionException e) {
          e.printStackTrace();
        }

        return "task " + backgroundTask.id;
      }
    });

    // display the data after they are loaded
    CompletableFuture<Void> future = completableFuture.thenAccept(x -> {

      System.out.println( "Background task finished:" + x);

    });

  }


  public static void main(String[] args) {

    // runnable which invokes the background loader every second
    Runnable trigger = new Runnable() {

      int taskId = 0;

      public void run() {

        loadInBackground( taskId++);

      }
    };

    // create scheduler
    ScheduledExecutorService scheduler = Executors.newScheduledThreadPool(1);
    ScheduledFuture<?> beeperHandle = scheduler.scheduleAtFixedRate(trigger, 0, 1, TimeUnit.SECONDS);

    // cancel the scheudler and the application after 10 seconds
    scheduler.schedule(() -> beeperHandle.cancel(true), 10, TimeUnit.SECONDS);

    try {
      beeperHandle.get();
    } catch (Throwable th) {
    }

    System.out.println( "Cancelled");
    System.exit(0);
  }

}

결과는 다음과 같습니다.

Start #0: Thread[ForkJoinPool.commonPool-worker-2,5,main]
Start #1: Thread[ForkJoinPool.commonPool-worker-4,5,main]
Start #2: Thread[ForkJoinPool.commonPool-worker-6,5,main]
Finish #0: Thread[ForkJoinPool.commonPool-worker-2,5,main]
Background task finished:task 0
Finish #1: Thread[ForkJoinPool.commonPool-worker-4,5,main]
Background task finished:task 1
Start #3: Thread[ForkJoinPool.commonPool-worker-4,5,main]
Finish #2: Thread[ForkJoinPool.commonPool-worker-6,5,main]
Background task finished:task 2
Start #4: Thread[ForkJoinPool.commonPool-worker-6,5,main]
Start #5: Thread[ForkJoinPool.commonPool-worker-2,5,main]
Finish #3: Thread[ForkJoinPool.commonPool-worker-4,5,main]
Background task finished:task 3
Start #6: Thread[ForkJoinPool.commonPool-worker-4,5,main]
Finish #4: Thread[ForkJoinPool.commonPool-worker-6,5,main]
Background task finished:task 4
Finish #5: Thread[ForkJoinPool.commonPool-worker-2,5,main]
Background task finished:task 5
Start #7: Thread[ForkJoinPool.commonPool-worker-2,5,main]
Finish #6: Thread[ForkJoinPool.commonPool-worker-4,5,main]
Start #8: Thread[ForkJoinPool.commonPool-worker-6,5,main]
Background task finished:task 6
Start #9: Thread[ForkJoinPool.commonPool-worker-4,5,main]
Finish #7: Thread[ForkJoinPool.commonPool-worker-2,5,main]
Background task finished:task 7
Start #10: Thread[ForkJoinPool.commonPool-worker-2,5,main]
Finish #8: Thread[ForkJoinPool.commonPool-worker-6,5,main]
Background task finished:task 8
Cancelled

목표는 # 0이 여전히 실행 중이므로 예를 들어 # 1과 # 2를 건너 뛰는 것입니다.

문제

차단 메커니즘을 어디에서 올바르게 설정합니까? 동기화를 사용해야합니까? 아니면 AtomicBoolean? 그렇다면 get()방법 내부 또는 다른 곳에 있어야 합니까?



답변

작업을 실행할 스레드 풀이 이미 있습니다. 반드시 다른 비동기 실행기에서 작업을 실행하는 것이 반드시 복잡하지는 않습니다 ( ForkJoinPool을 사용할 때 CompletableFuture)

간단하게:

public static void loadInBackground(int taskId) {
    // create the loading task
    BackgroundTask backgroundTask = new BackgroundTask(taskId);
    // No need to run in async, as it already in executor
    backgroundTask.run();
}

당신이 그것을 호출 할 때 ScheduledExecutorService를 단 하나의 작업이 시간에 실행되도록해야합니다 scheduleAtFixedRate

지정된 초기 지연 이후에 지정된 주기로 활성화되는 주기적 조치를 작성하고 실행합니다. 즉, initialDelay, initialDelay + 마침표, initialDelay + 2 * 기간 등이 실행 된 후에 실행이 시작됩니다. 작업 실행에서 예외가 발생하면 후속 실행이 억제됩니다. 그렇지 않으면 작업은 실행기의 취소 또는 종료를 통해서만 종료됩니다.이 작업의 실행이 기간보다 오래 걸리면 후속 실행이 늦게 시작될 수 있지만 동시에 실행되지는 않습니다 .


답변

요구 사항으로 다음을 고려하십시오.

  • 백그라운드에서 데이터를로드해야합니다
  • 로드 후 데이터가 표시되어야합니다
  • 데이터가로드되는 동안 더 이상 요청을 수락하지 않아야합니다.
  • 데이터가로드되는 동안 요청이있는 경우 특정 시간 초과 (예 : 5 초) 후에 다른로드를 예약해야합니다.

이 솔루션은 빌드에 기초 CA는 Executors.newSingleThreadExecutor(), CompletableFutureLinkedBlockingQueue:

public class SingleThreadedLoader {

  private static class BackgroundTask extends CompletableFuture<String> {

    private final String query;

    private BackgroundTask(String query) {
      this.query = query;
    }

    public String getQuery() {
      return query;
    }
  }

  private final BlockingQueue<BackgroundTask> tasks = new LinkedBlockingQueue<>();
  // while data are loaded no further requests should be accepted
  private final Executor executor = Executors.newSingleThreadExecutor();

  private final int delaySeconds;

  private AtomicReference<Instant> lastExecution = new AtomicReference<>(Instant.EPOCH);

  public SingleThreadedLoader(int delaySeconds) {
    this.delaySeconds = delaySeconds;
    setupLoading();
  }

  public BackgroundTask loadInBackground(String query) {
    log("Enqueued query " + query);
    BackgroundTask task = new BackgroundTask(query);
    tasks.add(task);
    return task;
  }

  private void setupLoading() {
    // data should be loaded in background
    executor.execute(() -> {
      while (true) {
        try {
          // if there were requests while the data were loaded
          // another loading should be scheduled after a certain timeout (e. g. 5 seconds)
          Instant prev = lastExecution.get();
          long delay = Duration.between(prev, Instant.now()).toSeconds();
          if (delay < delaySeconds) {
            log("Waiting for 5 seconds before next data loading");
            TimeUnit.SECONDS.sleep(delaySeconds - delay);
          }
          BackgroundTask task = tasks.take();
          try {
            String query = task.getQuery();
            String data = loadData(query);
            task.complete(data);
          } catch (Exception e) {
            task.completeExceptionally(e);
          }
          lastExecution.set(Instant.now());
        } catch (InterruptedException e) {
          log(e.getMessage());
          return;
        }
      }
    });
  }

  private String loadData(String query) {
    try {
      log("Loading data for " + query);
      TimeUnit.SECONDS.sleep(2);
      log("Loaded data for " + query);
      return "Result " + query;
    } catch (InterruptedException e) {
      throw new RuntimeException(e);
    }
  }

  private static void log(String str) {
    String time = LocalTime.now().truncatedTo(ChronoUnit.SECONDS).format(DateTimeFormatter.ISO_TIME);
    String thread = Thread.currentThread().getName();
    System.out.println(time + ' ' + thread + ": " + str);
  }

  public static void main(String[] args) throws Exception {
    SingleThreadedLoader loader = new SingleThreadedLoader(5);
    // after the loading the data should be displayed
    loader.loadInBackground("1").thenAccept(SingleThreadedLoader::log);
    loader.loadInBackground("2").thenAccept(SingleThreadedLoader::log);
    loader.loadInBackground("3").thenAccept(SingleThreadedLoader::log);

    log("Do another work in the main thread");

    TimeUnit.SECONDS.sleep(30);
  }
}

실행 후 stdout은 다음과 같은 출력을 갖습니다.

10:29:26 main: Enqueued query 1
10:29:26 pool-1-thread-1: Loading data for 1
10:29:26 main: Enqueued query 2
10:29:26 main: Enqueued query 3
10:29:26 main: Do another work in the main thread
10:29:28 pool-1-thread-1: Loaded data for 1
10:29:28 pool-1-thread-1: Result 1
10:29:28 pool-1-thread-1: Waiting for 5 seconds before next data loading
10:29:33 pool-1-thread-1: Loading data for 2
10:29:36 pool-1-thread-1: Loaded data for 2
10:29:36 pool-1-thread-1: Result 2
10:29:36 pool-1-thread-1: Waiting for 5 seconds before next data loading
10:29:41 pool-1-thread-1: Loading data for 3
10:29:43 pool-1-thread-1: Loaded data for 3
10:29:43 pool-1-thread-1: Result 3


답변

나는 간단한 lock () 및 unlock () 메소드로 작업을 실행하는 카운터 역할을하는 AtomicInteger를 추가하여 출력을 얻은 원래 코드로 약간 변경했습니다.

Start #0: Thread[ForkJoinPool.commonPool-worker-2,5,main]
background task cancelled  1
background task cancelled  2
Finish #0: Thread[ForkJoinPool.commonPool-worker-2,5,main]
Background task finished:task 0
Start #3: Thread[ForkJoinPool.commonPool-worker-2,5,main]
background task cancelled  4
Finish #3: Thread[ForkJoinPool.commonPool-worker-2,5,main]
background task cancelled  5
Background task finished:task 3
Start #6: Thread[ForkJoinPool.commonPool-worker-3,5,main]
background task cancelled  7
Finish #6: Thread[ForkJoinPool.commonPool-worker-3,5,main]
background task cancelled  8
Background task finished:task 6
Start #9: Thread[ForkJoinPool.commonPool-worker-2,5,main]
background task cancelled  10
Cancelled

귀하의 작업에 대한 나의 해결책은 다음과 같습니다.

public class LoadInBackgroundExample {
    //Added new exception
    public static class AlreadyIsRunningException extends RuntimeException {
        long taskId;

        public AlreadyIsRunningException(String message, long taskId) {
            super(message);
            this.taskId = taskId;
        }

        public long getTaskId() {
            return taskId;
        }

        public void setTaskId(long taskId) {
            this.taskId = taskId;
        }
    }


    /**
     * A simple background task which should perform the data loading operation. In this minimal example it simply invokes Thread.sleep
     */
    public static class BackgroundTask implements Runnable {


        //this atomicInteger acts as a global lock counter for BackgroundTask objects
        private static AtomicInteger counter = new AtomicInteger(0);


        private int id;

        public BackgroundTask(int id) {
            this.id = id;
        }

        private void unlock() {
            counter.decrementAndGet();
        }

        private void lock() {
            //we need to check this way to avoid some unlucky timing between threads
            int lockValue = counter.incrementAndGet();
            //if we got counter different than 1 that means that some other task is already running (it has already acquired the lock)
            if (lockValue != 1) {
                //rollback our check
                counter.decrementAndGet();
                //throw an exception
                throw new AlreadyIsRunningException("Some other task already is running", id);
            }
        }

        /**
         * Sleep for a given amount of time to simulate loading.
         */
        @Override
        public void run() {
            //Check if we can acquire lock

            lock();
            //we have a lock to
            try {
                System.out.println("Start #" + id + ": " + Thread.currentThread());
                long sleepTime = 2000;
                Thread.sleep(sleepTime);
            } catch (InterruptedException e) {
                e.printStackTrace();
            } finally {
                System.out.println("Finish #" + id + ": " + Thread.currentThread());
                unlock();
            }
        }
    }

    /**
     * CompletableFuture which simulates loading and showing data.
     *
     * @param taskId Identifier of the current task
     */


    public static void loadInBackground(int taskId) {
        // create the loading task
        BackgroundTask backgroundTask = new BackgroundTask(taskId);
        // "load" the data asynchronously
        CompletableFuture<String> completableFuture = CompletableFuture.supplyAsync(new Supplier<String>() {
            @Override
            public String get() {
                CompletableFuture<Void> future = CompletableFuture.runAsync(backgroundTask);

                try {
                    future.get();
                } catch (ExecutionException e) {
                    if (e.getCause() instanceof AlreadyIsRunningException) {
                        System.out.println("background task cancelled  " + ((AlreadyIsRunningException) e.getCause()).getTaskId());
                        throw (AlreadyIsRunningException) e.getCause();
                    }
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
                return "task " + backgroundTask.id;
            }
        });
        // display the data after they are loaded
        CompletableFuture<Void> future = completableFuture.thenAccept(x -> {
            System.out.println("Background task finished:" + x);
        });
    }

    ArrayList<BackgroundTask> backgroundTasks = new ArrayList<>();


    public static void main(String[] args) {
        // runnable which invokes the background loader every second
        Runnable trigger = new Runnable() {
            int taskId = 0;

            public void run() {
                loadInBackground(taskId++);
            }
        };

        // create scheduler
        ScheduledExecutorService scheduler = Executors.newScheduledThreadPool(1);
        ScheduledFuture<?> beeperHandle = scheduler.scheduleAtFixedRate(trigger, 0, 1, TimeUnit.SECONDS);

        // cancel the scheudler and the application after 10 seconds
        scheduler.schedule(() -> beeperHandle.cancel(true), 10, TimeUnit.SECONDS);

        try {
            beeperHandle.get();
        } catch (Throwable th) {
        }

        System.out.println("Cancelled");
        System.exit(0);
    }

최신 정보

lock () 및 unlock () 메소드를보다 간단한 형식으로 변경했습니다.

private static  AtomicBoolean atomicBoolean = new AtomicBoolean(false);
        private void unlock() {
            atomicBoolean.set(false);
        }
        private void lock() {
            //if 'changed' is false that means some other task is already running
            boolean changed = atomicBoolean.compareAndSet(false,true);
            if (!changed) {
                throw new AlreadyIsRunningException("Some other task  is already running", id);
            }
        }


답변

이해하면 동시에 여러 가지 작업을 백그라운드로 수행 할 수 있습니다. 이러한 작업은 정확히 동일한 작업을 수행하므로 작업을 병렬로 실행하지 않으려면 작업을 완료하고 결과를 다른 작업과 공유하기 위해 하나의 작업이 필요합니다. 따라서 CompletableFuture동시에 10을 얻는다면 그들 중 하나가 ‘reload’를 db로 호출하고 실행 결과를 다른 사람들과 공유하는 방식으로CompletableFuture 결과가 정상적으로 완료 합니다. 나는 이것을

목표는 # 0이 여전히 실행 중이므로 예를 들어 # 1과 # 2를 건너 뛰는 것입니다.

로드 후 데이터가 표시되어야합니다

내 추측이 맞다면 내 해결책을 시도해 볼 수 있습니다.

나는 작업 사이에 어떤 종류의 부모-자식 관계가 있습니다. 부모 과제는 실제로 업무를 수행하고 자녀에게 얻은 결과를 공유하는 것입니다. 자식 작업은 부모 작업이 여전히 실행되는 동안 추가 된 작업이며 자식 작업은 부모 작업이 실행을 마칠 때까지 기다립니다. 부모 과제의 결과가 여전히 ‘신선한’상태이기 때문에 각 자녀에게 복사되어 모든 자녀가 미래를 완성합니다.

public class BackgroundService {


    public static class BackgroundJob implements Callable<String> {

        private static BackgroundJob ROOT_JOB = null;

        private synchronized static void addBackgroundJob(BackgroundJob backgroundJob) {
            if (ROOT_JOB != null) {
                ROOT_JOB.addChild(backgroundJob);
            } else {
                System.out.println();
                System.out.println(Thread.currentThread().getName() + " RUNNING ROOT TASK-" + backgroundJob.jobId);
                ROOT_JOB = backgroundJob;
            }
        }

        private synchronized static void unlock() {
            ROOT_JOB = null;
        }

        private final int jobId;
        private List<BackgroundJob> children = new ArrayList<>();
        private BackgroundJob parent;
        private String providedResultFromParent = null;


        public BackgroundJob(int jobId) {
            this.jobId = jobId;
        }

        private  void addChild(BackgroundJob backgroundJob) {
            backgroundJob.parent = this;
            this.children.add(backgroundJob);
        }

        @Override
        public String call() throws Exception {
            addBackgroundJob(this);
            if (parent == null) {
                String result = logic();

                synchronized (ROOT_JOB) {
                    for (final BackgroundJob backgroundJob : children) {
                        backgroundJob.providedResultFromParent = result;
                        synchronized (backgroundJob) {
                            backgroundJob.notify();
                        }
                    }
                    unlock();
                }

                return "\t\tROOT task" + jobId + "'s " + result;
            } else {
                synchronized (this) {
                    System.out.println(Thread.currentThread().getName() + "\t\tskipping task-" + jobId + " and waiting running task-" + parent.jobId + " to finish");
                    this.wait();
                }
                return "\t\t\t\ttask-" + jobId + "'s " + providedResultFromParent;
            }
        }

        private String logic() throws InterruptedException {
            Thread.sleep(2000);
            return (int) (Math.random() * 1000) + " ";
        }
    }

    public static void main(String[] args) throws InterruptedException, ExecutionException {
        AtomicInteger atomicInteger = new AtomicInteger();
        ExecutorService pool = Executors.newCachedThreadPool();
        Supplier<String> job = () -> {
            int taskId = atomicInteger.incrementAndGet();
            BackgroundJob backgroundJob = new BackgroundJob(taskId);
            try {
                return backgroundJob.call();
            } catch (Exception e) {
                e.printStackTrace();
            }
            return "finished " + taskId;
        };

        for (int i = 100; i > 0; i--) {
            CompletableFuture.supplyAsync(job, pool).thenAccept(s -> System.out.println(Thread.currentThread().getName()+" "+ s + " result is readable"));
            Thread.sleep((long) (Math.random() * 500));
        }

        pool.awaitTermination(Long.MAX_VALUE, TimeUnit.DAYS);
        pool.shutdown();
    }

그리고 출력은 다음과 같습니다.

pool-1-thread-1 RUNNING ROOT TASK-1
pool-1-thread-2     skipping task-2 and waiting running task-1 to finish
pool-1-thread-3     skipping task-3 and waiting running task-1 to finish
pool-1-thread-4     skipping task-4 and waiting running task-1 to finish
pool-1-thread-5     skipping task-5 and waiting running task-1 to finish
pool-1-thread-6     skipping task-6 and waiting running task-1 to finish
pool-1-thread-7     skipping task-7 and waiting running task-1 to finish
pool-1-thread-8     skipping task-8 and waiting running task-1 to finish
pool-1-thread-3                 task-3's 165  result is readable
pool-1-thread-8                 task-8's 165  result is readable
pool-1-thread-6                 task-6's 165  result is readable
pool-1-thread-5                 task-5's 165  result is readable
pool-1-thread-7                 task-7's 165  result is readable
pool-1-thread-2                 task-2's 165  result is readable
pool-1-thread-1         ROOT task1's 165  result is readable
pool-1-thread-4                 task-4's 165  result is readable

pool-1-thread-4 RUNNING ROOT TASK-9
pool-1-thread-1     skipping task-10 and waiting running task-9 to finish
pool-1-thread-2     skipping task-11 and waiting running task-9 to finish
pool-1-thread-7     skipping task-12 and waiting running task-9 to finish
pool-1-thread-5     skipping task-13 and waiting running task-9 to finish
pool-1-thread-8     skipping task-14 and waiting running task-9 to finish
pool-1-thread-6     skipping task-15 and waiting running task-9 to finish
pool-1-thread-3     skipping task-16 and waiting running task-9 to finish
pool-1-thread-9     skipping task-17 and waiting running task-9 to finish
pool-1-thread-10        skipping task-18 and waiting running task-9 to finish
pool-1-thread-1                 task-10's 370  result is readable
pool-1-thread-10                task-18's 370  result is readable
pool-1-thread-4         ROOT task9's 370  result is readable
pool-1-thread-9                 task-17's 370  result is readable
pool-1-thread-7                 task-12's 370  result is readable
pool-1-thread-6                 task-15's 370  result is readable
pool-1-thread-8                 task-14's 370  result is readable
pool-1-thread-2                 task-11's 370  result is readable
pool-1-thread-3                 task-16's 370  result is readable
pool-1-thread-5                 task-13's 370  result is readable

pool-1-thread-5 RUNNING ROOT TASK-19
pool-1-thread-3     skipping task-20 and waiting running task-19 to finish
pool-1-thread-2     skipping task-21 and waiting running task-19 to finish
pool-1-thread-8     skipping task-22 and waiting running task-19 to finish
pool-1-thread-6     skipping task-23 and waiting running task-19 to finish
pool-1-thread-7     skipping task-24 and waiting running task-19 to finish
pool-1-thread-9     skipping task-25 and waiting running task-19 to finish
pool-1-thread-4     skipping task-26 and waiting running task-19 to finish
pool-1-thread-10        skipping task-27 and waiting running task-19 to finish
pool-1-thread-1     skipping task-28 and waiting running task-19 to finish
pool-1-thread-5         ROOT task19's 574  result is readable
pool-1-thread-8                 task-22's 574  result is readable
pool-1-thread-4                 task-26's 574  result is readable
pool-1-thread-7                 task-24's 574  result is readable
pool-1-thread-6                 task-23's 574  result is readable
pool-1-thread-3                 task-20's 574  result is readable
pool-1-thread-9                 task-25's 574  result is readable
pool-1-thread-2                 task-21's 574  result is readable
pool-1-thread-1                 task-28's 574  result is readable
pool-1-thread-10                task-27's 574  result is readable

pool-1-thread-10 RUNNING ROOT TASK-29
pool-1-thread-1     skipping task-30 and waiting running task-29 to finish
pool-1-thread-2     skipping task-31 and waiting running task-29 to finish
pool-1-thread-9     skipping task-32 and waiting running task-29 to finish
pool-1-thread-3     skipping task-33 and waiting running task-29 to finish
pool-1-thread-6     skipping task-34 and waiting running task-29 to finish
pool-1-thread-7     skipping task-35 and waiting running task-29 to finish
pool-1-thread-4     skipping task-36 and waiting running task-29 to finish
pool-1-thread-8     skipping task-37 and waiting running task-29 to finish
pool-1-thread-5     skipping task-38 and waiting running task-29 to finish
pool-1-thread-11        skipping task-39 and waiting running task-29 to finish
pool-1-thread-1                 task-30's 230  result is readable
pool-1-thread-11                task-39's 230  result is readable
pool-1-thread-8                 task-37's 230  result is readable
pool-1-thread-5                 task-38's 230  result is readable
pool-1-thread-4                 task-36's 230  result is readable
pool-1-thread-7                 task-35's 230  result is readable

pool-1-thread-12 RUNNING ROOT TASK-40
pool-1-thread-6                 task-34's 230  result is readable
pool-1-thread-10        ROOT task29's 230  result is readable
pool-1-thread-3                 task-33's 230  result is readable
pool-1-thread-9                 task-32's 230  result is readable
pool-1-thread-2                 task-31's 230  result is readable
pool-1-thread-2     skipping task-41 and waiting running task-40 to finish
pool-1-thread-9     skipping task-42 and waiting running task-40 to finish
pool-1-thread-3     skipping task-43 and waiting running task-40 to finish
pool-1-thread-10        skipping task-44 and waiting running task-40 to finish
pool-1-thread-6     skipping task-45 and waiting running task-40 to finish
pool-1-thread-7     skipping task-46 and waiting running task-40 to finish
pool-1-thread-2                 task-41's 282  result is readable
pool-1-thread-10                task-44's 282  result is readable
pool-1-thread-6                 task-45's 282  result is readable
pool-1-thread-7                 task-46's 282  result is readable
pool-1-thread-3                 task-43's 282  result is readable
pool-1-thread-9                 task-42's 282  result is readable
pool-1-thread-12        ROOT task40's 282  result is readable

pool-1-thread-12 RUNNING ROOT TASK-47
pool-1-thread-9     skipping task-48 and waiting running task-47 to finish
pool-1-thread-3     skipping task-49 and waiting running task-47 to finish
pool-1-thread-7     skipping task-50 and waiting running task-47 to finish
pool-1-thread-6     skipping task-51 and waiting running task-47 to finish
pool-1-thread-10        skipping task-52 and waiting running task-47 to finish
pool-1-thread-2     skipping task-53 and waiting running task-47 to finish
pool-1-thread-12        ROOT task47's 871  result is readable
pool-1-thread-10                task-52's 871  result is readable
pool-1-thread-2                 task-53's 871  result is readable
pool-1-thread-3                 task-49's 871  result is readable
pool-1-thread-6                 task-51's 871  result is readable
pool-1-thread-7                 task-50's 871  result is readable
pool-1-thread-9                 task-48's 871  result is readable

pool-1-thread-9 RUNNING ROOT TASK-54
pool-1-thread-7     skipping task-55 and waiting running task-54 to finish
pool-1-thread-6     skipping task-56 and waiting running task-54 to finish
pool-1-thread-3     skipping task-57 and waiting running task-54 to finish
pool-1-thread-2     skipping task-58 and waiting running task-54 to finish
pool-1-thread-10        skipping task-59 and waiting running task-54 to finish
pool-1-thread-12        skipping task-60 and waiting running task-54 to finish
pool-1-thread-4     skipping task-61 and waiting running task-54 to finish
pool-1-thread-5     skipping task-62 and waiting running task-54 to finish
pool-1-thread-9         ROOT task54's 345  result is readable
pool-1-thread-2                 task-58's 345  result is readable
pool-1-thread-5                 task-62's 345  result is readable
pool-1-thread-7                 task-55's 345  result is readable
pool-1-thread-10                task-59's 345  result is readable
pool-1-thread-6                 task-56's 345  result is readable
pool-1-thread-3                 task-57's 345  result is readable
pool-1-thread-4                 task-61's 345  result is readable
pool-1-thread-12                task-60's 345  result is readable

pool-1-thread-12 RUNNING ROOT TASK-63
pool-1-thread-4     skipping task-64 and waiting running task-63 to finish
pool-1-thread-3     skipping task-65 and waiting running task-63 to finish
pool-1-thread-6     skipping task-66 and waiting running task-63 to finish
pool-1-thread-10        skipping task-67 and waiting running task-63 to finish
pool-1-thread-7     skipping task-68 and waiting running task-63 to finish
pool-1-thread-5     skipping task-69 and waiting running task-63 to finish
pool-1-thread-2     skipping task-70 and waiting running task-63 to finish
pool-1-thread-12        ROOT task63's 670  result is readable
pool-1-thread-2                 task-70's 670  result is readable
pool-1-thread-5                 task-69's 670  result is readable
pool-1-thread-7                 task-68's 670  result is readable
pool-1-thread-10                task-67's 670  result is readable
pool-1-thread-6                 task-66's 670  result is readable
pool-1-thread-3                 task-65's 670  result is readable
pool-1-thread-4                 task-64's 670  result is readable

pool-1-thread-4 RUNNING ROOT TASK-71
pool-1-thread-3     skipping task-72 and waiting running task-71 to finish
pool-1-thread-6     skipping task-73 and waiting running task-71 to finish
pool-1-thread-10        skipping task-74 and waiting running task-71 to finish
pool-1-thread-7     skipping task-75 and waiting running task-71 to finish
pool-1-thread-5     skipping task-76 and waiting running task-71 to finish
pool-1-thread-2     skipping task-77 and waiting running task-71 to finish
pool-1-thread-12        skipping task-78 and waiting running task-71 to finish
pool-1-thread-9     skipping task-79 and waiting running task-71 to finish
pool-1-thread-8     skipping task-80 and waiting running task-71 to finish
pool-1-thread-4         ROOT task71's 445  result is readable
pool-1-thread-6                 task-73's 445  result is readable
pool-1-thread-9                 task-79's 445  result is readable
pool-1-thread-3                 task-72's 445  result is readable
pool-1-thread-8                 task-80's 445  result is readable
pool-1-thread-12                task-78's 445  result is readable
pool-1-thread-5                 task-76's 445  result is readable
pool-1-thread-10                task-74's 445  result is readable
pool-1-thread-2                 task-77's 445  result is readable
pool-1-thread-7                 task-75's 445  result is readable

pool-1-thread-7 RUNNING ROOT TASK-81
pool-1-thread-2     skipping task-82 and waiting running task-81 to finish
pool-1-thread-10        skipping task-83 and waiting running task-81 to finish
pool-1-thread-5     skipping task-84 and waiting running task-81 to finish
pool-1-thread-12        skipping task-85 and waiting running task-81 to finish
pool-1-thread-8     skipping task-86 and waiting running task-81 to finish
pool-1-thread-3     skipping task-87 and waiting running task-81 to finish
pool-1-thread-9     skipping task-88 and waiting running task-81 to finish
pool-1-thread-6     skipping task-89 and waiting running task-81 to finish
pool-1-thread-7         ROOT task81's 141  result is readable
pool-1-thread-6                 task-89's 141  result is readable
pool-1-thread-9                 task-88's 141  result is readable
pool-1-thread-3                 task-87's 141  result is readable
pool-1-thread-10                task-83's 141  result is readable
pool-1-thread-5                 task-84's 141  result is readable
pool-1-thread-12                task-85's 141  result is readable
pool-1-thread-8                 task-86's 141  result is readable
pool-1-thread-2                 task-82's 141  result is readable

pool-1-thread-2 RUNNING ROOT TASK-90
pool-1-thread-8     skipping task-91 and waiting running task-90 to finish
pool-1-thread-12        skipping task-92 and waiting running task-90 to finish
pool-1-thread-5     skipping task-93 and waiting running task-90 to finish
pool-1-thread-10        skipping task-94 and waiting running task-90 to finish
pool-1-thread-3     skipping task-95 and waiting running task-90 to finish
pool-1-thread-9     skipping task-96 and waiting running task-90 to finish
pool-1-thread-6     skipping task-97 and waiting running task-90 to finish
pool-1-thread-7     skipping task-98 and waiting running task-90 to finish
pool-1-thread-4     skipping task-99 and waiting running task-90 to finish
pool-1-thread-11        skipping task-100 and waiting running task-90 to finish
pool-1-thread-2         ROOT task90's 321  result is readable
pool-1-thread-3                 task-95's 321  result is readable
pool-1-thread-7                 task-98's 321  result is readable
pool-1-thread-8                 task-91's 321  result is readable
pool-1-thread-11                task-100's 321  result is readable
pool-1-thread-4                 task-99's 321  result is readable
pool-1-thread-5                 task-93's 321  result is readable
pool-1-thread-9                 task-96's 321  result is readable
pool-1-thread-12                task-92's 321  result is readable
pool-1-thread-10                task-94's 321  result is readable
pool-1-thread-6                 task-97's 321  result is readable


답변

하나의 단일 액세스 스레드 만 원하면 간단한 동기화 작업이 수행됩니다 …

산출:

Start #2: Thread[ForkJoinPool.commonPool-worker-6,5,main]
Finish #0: Thread[ForkJoinPool.commonPool-worker-2,5,main]
Background task finished:task 0 finished getting data...
Start #3: Thread[ForkJoinPool.commonPool-worker-2,5,main]
Finish #2: Thread[ForkJoinPool.commonPool-worker-6,5,main]
Start #4: Thread[ForkJoinPool.commonPool-worker-6,5,main]
Background task finished:task 2 finished getting data...
Finish #1: Thread[ForkJoinPool.commonPool-worker-4,5,main]
Background task finished:task 1 finished getting data...
Start #6: Thread[ForkJoinPool.commonPool-worker-3,5,main]
Finish #4: Thread[ForkJoinPool.commonPool-worker-6,5,main]
Start #5: Thread[ForkJoinPool.commonPool-worker-6,5,main]
Background task finished:task 4 finished getting data...
Finish #3: Thread[ForkJoinPool.commonPool-worker-2,5,main]
Start #7: Thread[ForkJoinPool.commonPool-worker-2,5,main]
Background task finished:task 3 finished getting data...
Cancelled

암호:

package queue;

import java.util.concurrent.CompletableFuture;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.Executors;
import java.util.concurrent.ScheduledExecutorService;
import java.util.concurrent.ScheduledFuture;
import java.util.concurrent.TimeUnit;
import java.util.function.Supplier;

public class LoadInBackgroundExample {

    public static class SyncronizedBackend {
        public synchronized String getData() {

            long sleepTime = 2000;
            try {
                Thread.sleep(sleepTime);
            } catch (InterruptedException e) {
                // TODO Auto-generated catch block
                e.printStackTrace();
            }

            return new String("finished getting data...");
        }
    }

    /**
     * A simple background task which should perform the data loading operation. In
     * this minimal example it simply invokes Thread.sleep
     */
    public static class BackgroundTask implements Runnable {

        private int id;
        private SyncronizedBackend syncronizedBackend;
        private String result;

        public BackgroundTask(SyncronizedBackend syncronizedBackend, int id) {
            this.syncronizedBackend = syncronizedBackend;
            this.id = id;
        }

        /**
         * Sleep for a given amount of time to simulate loading.
         */
        @Override
        public void run() {

            System.out.println("Start #" + id + ": " + Thread.currentThread());

            result = this.syncronizedBackend.getData();

            System.out.println("Finish #" + id + ": " + Thread.currentThread());
        }

        public String getResult() {
            return result;
        }
    }

    /**
     * CompletableFuture which simulates loading and showing data.
     * @param syncronizedBackend
     *
     * @param taskId Identifier of the current task
     */
    public static void loadInBackground(SyncronizedBackend syncronizedBackend, int taskId) {

        // create the loading task
        BackgroundTask backgroundTask = new BackgroundTask(syncronizedBackend, taskId);

        // "load" the data asynchronously
        CompletableFuture<String> completableFuture = CompletableFuture.supplyAsync(new Supplier<String>() {

            @Override
            public String get() {

                CompletableFuture<Void> future = CompletableFuture.runAsync(backgroundTask);

                try {

                    future.get();

                } catch (InterruptedException | ExecutionException e) {
                    e.printStackTrace();
                }

                return "task " + backgroundTask.id + " " + backgroundTask.getResult();
            }
        });

        // display the data after they are loaded
        CompletableFuture<Void> future = completableFuture.thenAccept(x -> {

            System.out.println("Background task finished:" + x);

        });

    }

    public static void main(String[] args) {

        SyncronizedBackend syncronizedBackend = new SyncronizedBackend();

        // runnable which invokes the background loader every second
        Runnable trigger = new Runnable() {

            int taskId = 0;

            public void run() {

                loadInBackground(syncronizedBackend, taskId++);

            }
        };

        // create scheduler
        ScheduledExecutorService scheduler = Executors.newScheduledThreadPool(1);
        ScheduledFuture<?> beeperHandle = scheduler.scheduleAtFixedRate(trigger, 0, 1, TimeUnit.SECONDS);

        // cancel the scheudler and the application after 10 seconds
        scheduler.schedule(() -> beeperHandle.cancel(true), 10, TimeUnit.SECONDS);

        try {
            beeperHandle.get();
        } catch (Throwable th) {
        }

        System.out.println("Cancelled");
        System.exit(0);
    }

}


답변

I 클래스, 스레드 듀얼 스위치를 사용하여 솔루션을 참조 시도 BackgroundTaskDualSwitch, 그것은 시뮬레이트 로딩이를 사용 CompletableFuture. 아이디어는 현재 실행중인 작업이 완료 될 때까지 두 번째 작업이 대기하도록하는 것입니다 (변경 사항 참조) BackgroundTask. 이를 통해 최대 하나의 작업 스레드가 실행 중이고 최대 하나의 작업 스레드가 대기 중입니다. 실행중인 작업이 완료 될 때까지 추가 요청을 건너 뛰고 다음 요청을 처리하기 위해 ‘자유’상태가됩니다.

public static class BackgroundTask extends Thread {

  private int id;
  private Thread pendingTask;

  public BackgroundTask(int id) {
    this.id = id;
  }

  public BackgroundTask(int id, Thread pendingTask) {
    this(id);
    this.pendingTask = pendingTask;
  }

  /**
   * Sleep for a given amount of time to simulate loading.
   */
  @Override
  public void run() {

    try {

      if (pendingTask != null && pendingTask.isAlive()) {
        pendingTask.join();
      }

      System.out.println("Start #" + id + ": " + Thread.currentThread());
      ...
    }
  }

public static class BackgroundTaskDualSwitch {

  private static BackgroundTask task1;
  private static BackgroundTask task2;

  public static synchronized boolean runTask(int taskId) {
    if (! isBusy(task1)) {
      if (isBusy(task2)) {
        task1 = new BackgroundTask(taskId, task2);
      } else {
        task1 = new BackgroundTask(taskId);
      }
      runAsync(task1);
      return true;
    } else if (! isBusy(task2)) {
      if (isBusy(task1)) {
        task2 = new BackgroundTask(taskId, task1);
      } else {
        task2 = new BackgroundTask(taskId);
      }
      runAsync(task2);
      return true;
    } else {
      return false; // SKIPPED
    }
  }

  private static void runAsync(BackgroundTask task) {
    CompletableFuture<String> completableFuture = CompletableFuture.supplyAsync(new Supplier<String>() {

      @Override
      public String get() {

        try {
          task.start();
          task.join();
        }
        catch (InterruptedException e) {
          e.printStackTrace();
        }
        return "task " + task.id;
      }
    });

    // display the data after they are loaded
    CompletableFuture<Void> future = completableFuture.thenAccept(x -> {

      System.out.println( "Background task finished:" + x);

    });
  }

  private static boolean isBusy(BackgroundTask task) {
    return task != null && task.isAlive();
  }
}

/**
 * Simulates loading and showing data.
 * @param taskId Identifier of the current task
 */
public static void loadInBackground(int taskId) {

  // create the loading task
  if (! BackgroundTaskDualSwitch.runTask(taskId)) {
    System.out.println( "Background task ignored:task " + taskId); // SKIPPED
  }
}
...

출력은 다음과 같습니다

Start #0: Thread[Thread-0,5,main]
Background task ignored:task 2
Finish #0: Thread[Thread-0,5,main]
Start #1: Thread[Thread-1,5,main]
Background task finished:task 0
Background task ignored:task 4
Finish #1: Thread[Thread-1,5,main]
Start #3: Thread[Thread-2,5,main]
Background task finished:task 1
Background task ignored:task 6
Finish #3: Thread[Thread-2,5,main]
Start #5: Thread[Thread-3,5,main]
Background task finished:task 3
Background task ignored:task 8
Finish #5: Thread[Thread-3,5,main]
Start #7: Thread[Thread-4,5,main]
Background task finished:task 5
Background task ignored:task 10
Finish #7: Thread[Thread-4,5,main]
Start #9: Thread[Thread-5,5,main]
Background task finished:task 7
Cancelled


답변

고가의 작업을 시작한 첫 번째 스레드는 결과를 콜백으로 알립니다. 이를 실행하려는 다른 스레드는 ExpensiveWork.notificables에 등록되므로 값 비싼 작업이 완료되면 작업을 수행 한 스레드가이를 통지합니다.

한편 스레드는 5 초마다 결과를 확인합니다.

public class ExpensiveWorkTest {

    private final static int THREADS = 20;
    private final static long THREAD_TIMEOUT = 5000L;

    @Test
    public void example() throws InterruptedException {
        ExpensiveWork<String> expensiveWork = new ExpensiveWorkImpl();
        ExecutorService service = Executors.newFixedThreadPool(THREADS);
        for(int i=0; i<THREADS;i++) {
            service.execute(() ->{
                Notificable<String> notificable = new NotificableImpl();
                expensiveWork.execute(notificable);
                while(notificable.getExpensiveResult() == null) {
                    try {
                        Thread.sleep(THREAD_TIMEOUT);
                    } catch (InterruptedException e) {}
                }
                System.out.println(Thread.currentThread().getName()+" has the message: "+notificable.getExpensiveResult());
            });
        }
        service.awaitTermination(60, TimeUnit.SECONDS);
    }

    public static abstract class ExpensiveWork<T> {

        private final AtomicBoolean runnning = new AtomicBoolean(false);
        private List<Notificable<T>> notificables = Collections.synchronizedList(new ArrayList<>());

        public void execute(Notificable<T> notificable) {
            String id = Thread.currentThread().getName();
            System.out.println("Loading data for "+id);
            notificables.add(notificable);
            if(!runnning.getAndSet(true)) {
                System.out.println("Running the expensive work "+id);
                T expensiveResult = expensiveWork();
                notificables.stream().forEach(n -> n.callback(expensiveResult));
            } else {
                System.out.println(id+" will receive the response later");
            }
        }

        protected abstract T expensiveWork();

    }

    public static class ExpensiveWorkImpl extends ExpensiveWork<String>{

        @Override
        public String expensiveWork() {
            try {
                Thread.sleep(5000);
            } catch (InterruptedException e) {}
            return "<Expensive result>";
        }

    }

    public static interface Notificable<T> {
        void callback(T expensiveResult);
        T getExpensiveResult();
    }

    public static class NotificableImpl implements Notificable<String> {

        private volatile String expensiveResult;

        @Override
        public void callback(String expensiveResult) {
            this.expensiveResult = expensiveResult;
        }

        @Override
        public String getExpensiveResult() {
            return expensiveResult;
        }

    }

}

그리고 이것은 출력입니다.

Loading data for pool-1-thread-6
Loading data for pool-1-thread-7
Loading data for pool-1-thread-9
pool-1-thread-9 will receive the response later
Loading data for pool-1-thread-8
pool-1-thread-8 will receive the response later
Loading data for pool-1-thread-1
Loading data for pool-1-thread-2
Loading data for pool-1-thread-12
pool-1-thread-12 will receive the response later
Loading data for pool-1-thread-3
pool-1-thread-3 will receive the response later
Loading data for pool-1-thread-4
Loading data for pool-1-thread-5
pool-1-thread-5 will receive the response later
pool-1-thread-4 will receive the response later
Loading data for pool-1-thread-14
pool-1-thread-14 will receive the response later
Loading data for pool-1-thread-13
Loading data for pool-1-thread-15
pool-1-thread-15 will receive the response later
pool-1-thread-2 will receive the response later
pool-1-thread-1 will receive the response later
Loading data for pool-1-thread-11
pool-1-thread-11 will receive the response later
Loading data for pool-1-thread-10
pool-1-thread-10 will receive the response later
pool-1-thread-7 will receive the response later
Running the expensive work pool-1-thread-6
Loading data for pool-1-thread-18
pool-1-thread-18 will receive the response later
Loading data for pool-1-thread-17
Loading data for pool-1-thread-16
pool-1-thread-16 will receive the response later
pool-1-thread-13 will receive the response later
Loading data for pool-1-thread-19
pool-1-thread-19 will receive the response later
pool-1-thread-17 will receive the response later
Loading data for pool-1-thread-20
pool-1-thread-20 will receive the response later
pool-1-thread-6 has the message: <Expensive result>
pool-1-thread-8 has the message: <Expensive result>
pool-1-thread-12 has the message: <Expensive result>
pool-1-thread-9 has the message: <Expensive result>
pool-1-thread-11 has the message: <Expensive result>
pool-1-thread-1 has the message: <Expensive result>
pool-1-thread-2 has the message: <Expensive result>
pool-1-thread-3 has the message: <Expensive result>
pool-1-thread-15 has the message: <Expensive result>
pool-1-thread-4 has the message: <Expensive result>
pool-1-thread-14 has the message: <Expensive result>
pool-1-thread-10 has the message: <Expensive result>
pool-1-thread-5 has the message: <Expensive result>
pool-1-thread-13 has the message: <Expensive result>
pool-1-thread-16 has the message: <Expensive result>
pool-1-thread-19 has the message: <Expensive result>
pool-1-thread-20 has the message: <Expensive result>
pool-1-thread-7 has the message: <Expensive result>
pool-1-thread-18 has the message: <Expensive result>
pool-1-thread-17 has the message: <Expensive result>