Spring Boot Batch：CPU性能优化示例

Question

In the answer for Configuring Spring Batch Steps in Parallel (Split) using Annotations an example ( FlowJobBuilderTests.java ) is given.

Question: is this answer really suitable for performance optimization? (I have some long running CPU consuming tasks to perform and need to use the available CPUs in an optimized way)

Answer 1

Example you are pointing to uses SimpleAsyncTaskExecutor .

Javadocs:

NOTE: This implementation does not reuse threads! Consider a thread-pooling TaskExecutor implementation instead, in particular for executing a large number of short-lived tasks.

So in order to make your CPU utilization more efficient, I would rather use ThreadPoolTaskExecutor or ConcurrentTaskExecutor .

As your tasks are CPU bound, I would configure thread pool size to amount of CPU cores on machine it will be running. So that your CPU cycles wouldn't be wasted on switching threads.

(Having bigger thread pool size would be suitable for IO bounds tasks, so that CPU wouldn't be waiting for IO operations)

Reaction on comment:

Now I better understand your concerns. Yes you can run steps in parallel. Here is example from my book about enterprise Spring :

  @Bean
  public Job prepareTeaJob(JobBuilderFactory jobBuilderFactory,
      @Qualifier("boilWaterStep") Step boilWaterStep,
      @Qualifier("addTeaStep") Step addTeaStep,
      @Qualifier("addSugarStep") Step addSugarStep,
      @Qualifier("addWaterStep") Step addWaterStep,
      TaskExecutor customTaskExecutor) {
    Job job = jobBuilderFactory.get("prepareTeaJob")
        .start(boilWaterStep)
        .split(customTaskExecutor)
        .add(new FlowBuilder<Flow>("addIngredientsSplit")
            .from(boilWaterStep).next(addTeaStep)
            .from(boilWaterStep).next(addSugarStep)
            .end())
        .next(addWaterStep)
        .end()
        .build();
    return job;
  }
}

From this example addTeaStep and addSugarStep will be executed in parallel, right after boilWaterStep .

Answer 2

Using

        Thread thread = Thread.currentThread();
        System.out.println(thread);

in the steps shows that the example really uses threads. Therefore the example is suitable.

Another good example that uses various threads is:

https://examples.javacodegeeks.com/enterprise-java/spring/batch/spring-batch-multithreading-example/ by Ashraf Sarhan.