Use cases for different types of concurrency constructs

1. ExecutorService

ExecutorService executor = Executors.newFixedThreadPool(50);

It is simple and easy to use. It hides low level details of ThreadPoolExecutor.

I prefer this one when number of Callable/Runnable tasks are small in number and piling of tasks in unbounded queue does not increase memory & degrade the performance of the system. If you have CPU/Memory constraints, I prefer to use ThreadPoolExecutor with capacity constraints & RejectedExecutionHandler to handle rejection of tasks.

1. CountDownLatch

CountDownLatch will be initialized with a given count. This count is decremented by calls to the countDown() method. Threads waiting for this count to reach zero can call one of the await() methods. Calling await() blocks the thread until the count reaches zero. This class enables a java thread to wait until other set of threads completes their tasks.

Use cases:

1. Achieving Maximum Parallelism: Sometimes we want to start a number of threads at the same time to achieve maximum parallelism
2. Wait N threads to completes before start execution
3. Deadlock detection.
4. ThreadPoolExecutor : It provides more control. If application is constrained by number of pending Runnable/Callable tasks, you can use bounded queue by setting the max capacity. Once the queue reaches maximum capacity, you can define RejectionHandler. Java provides four types of RejectedExecutionHandler policies.

1. `ThreadPoolExecutor.AbortPolicy`, the handler throws a runtime RejectedExecutionException upon rejection.

2. ThreadPoolExecutor.CallerRunsPolicy`, the thread that invokes execute itself runs the task. This provides a simple feedback control mechanism that will slow down the rate that new tasks are submitted.

3. In `ThreadPoolExecutor.DiscardPolicy`, a task that cannot be executed is simply dropped.

4. `ThreadPoolExecutor.DiscardOldestPolicy`, if the executor is not shut down, the task at the head of the work queue is dropped, and then execution is retried (which can fail again, causing this to be repeated.)

If you want to simulate CountDownLatch behaviour, you can use invokeAll() method.

1. One more mechanism you did not quote is ForkJoinPool

The ForkJoinPool was added to Java in Java 7. The ForkJoinPool is similar to the Java ExecutorService but with one difference. The ForkJoinPool makes it easy for tasks to split their work up into smaller tasks which are then submitted to the ForkJoinPool too. Task stealing happens in ForkJoinPool when free worker threads steal tasks from busy worker thread queue.

Java 8 has introduced one more API in ExecutorService to create work stealing pool. You don’t have to create RecursiveTask and RecursiveAction but still can use ForkJoinPool.

public static ExecutorService newWorkStealingPool()

Creates a work-stealing thread pool using all available processors as its target parallelism level.

By default, it will take number of CPU cores as parameter.

All these four mechanism are complimentary to each other. Depending on level of granularity you want to control, you have to chose right ones.