Customizing Thread Pools for Java parallelStream to Improve Performance and Isolation

In Java, parallelStream (introduced in Java 8) processes collection elements concurrently using the default ForkJoinPool.commonPool() , which may affect performance and lacks isolation, so a custom thread pool is often required.

1. Set system property : Adjust java.util.concurrent.ForkJoinPool.common.parallelism to change the parallelism of the shared pool.

public static void main(String[] args) throws InterruptedException {
    List
list = IntStream.range(1, 50).boxed().collect(Collectors.toList());
    list.parallelStream().forEach(t -> {
        System.out.println(t + ":" + Thread.currentThread().getName());
        try { TimeUnit.SECONDS.sleep(1); } catch (InterruptedException e) { e.printStackTrace(); }
    });
    TimeUnit.HOURS.sleep(1);
}

The default parallelism equals the number of CPU cores (e.g., 11 on the author's machine). You can query it with:

System.out.println("ForkJoinPool parallelism " + ForkJoinPool.getCommonPoolParallelism());

To increase it, set the property:

System.setProperty("java.util.concurrent.ForkJoinPool.common.parallelism", "20");
System.out.println("ForkJoinPool parallelism: " + ForkJoinPool.getCommonPoolParallelism());

Note: Changing the property only adjusts the shared pool's parallelism; it still does not provide isolation.

2. Run parallel operations in a custom ForkJoinPool : Create a dedicated pool and submit the parallel stream to it.

ForkJoinPool forkJoinPool = new ForkJoinPool(20);
forkJoinPool.submit(() -> list.parallelStream().forEach(t -> {
    System.out.println(t + ":" + Thread.currentThread().getName());
    try { TimeUnit.SECONDS.sleep(1); } catch (InterruptedException e) { e.printStackTrace(); }
}));
TimeUnit.HOURS.sleep(1);

Full example:

package com.renzhikeji.demo;

import java.util.List;
import java.util.concurrent.*;
import java.util.stream.IntStream;

public class JdkDemo {
    public static void main(String[] args) throws InterruptedException {
        List
list = IntStream.range(1, 50).boxed().toList();
        ForkJoinPool forkJoinPool = new ForkJoinPool(20);
        forkJoinPool.submit(() -> list.parallelStream().forEach(t -> {
            System.out.println(t + ":" + Thread.currentThread().getName());
            try { TimeUnit.SECONDS.sleep(1); } catch (InterruptedException e) { e.printStackTrace(); }
        }));
        TimeUnit.HOURS.sleep(1);
    }
}

The execution results show each element processed by a thread from the custom pool, confirming that the custom pool provides both higher parallelism and isolation.

Conclusion : Java's parallelStream may require a custom thread pool to improve performance and isolate failures; developers can either adjust the common pool's parallelism via a system property or create a dedicated ForkJoinPool and submit tasks to it.