multithreading - Java ThreadPool usage

Question

I'm trying to write a multithreaded web crawler.

My main entry class has the following code:

ExecutorService exec = Executors.newFixedThreadPool(numberOfCrawlers);
while(true){
    URL url = frontier.get();
    if(url == null)
         return;
exec.execute(new URLCrawler(this, url));
}

The URLCrawler fetches the specified URL, parses the HTML extracts links from it, and schedules unseen links back to frontier.

A frontier is a queue of uncrawled URLs. The problem is how to write the get() method. If the queue is empty, it should wait until any URLCrawlers finish and then try again. It should return null only when the queue is empty and there is no currently active URLCrawler.

My first idea was to use an AtomicInteger for counting current number of working URLCrawlers and an auxiliary object for notifyAll()/wait() calls. Each crawler on start increments the number of current working URLCrawlers, and on exit decrements it, and notify the object that it has completed.

But I read that notify()/notifyAll() and wait() are somewhat deprecated methods to do thread communication.

What should I use in this work pattern? It is similar to M producers and N consumers, the question is how to deal with exaustion of producers.

Answer 1

I am not sure I understand your design, but this may be a job for a Semaphore