如何并行执行多个模型而不是顺序执行？

Question

我有各种模型，我正在依次执行这些模型。我将有大约 200 个模型。

下面是我的代码，我在其中一个一个地执行这些模型

Map<String, List<ServerMetrics>> metricsHolder = new HashMap<String, List<ServerMetrics>>();

for (String alias : serverNames) {
    List<ClientsModel> modelMetadata = getModelAttributes();
    List<MachineInfo> machineInfo = getMachineInfo();

    List<ServerMetrics> metricsList = new ArrayList<ServerMetrics>();

    // calling model one by one sequentially
    // is there any way to make this multithreaded?
    // here modelMetadata size might be 100 or 200
    for (ClientsModel modelList : modelMetadata) {
        String modelValue = modelList.getModelValue();
        String modelId = String.valueOf(modelList.getModelId());

        // execute my model here and storing the metrics in the metricsList object
        ServerMetrics metrics = TestUtils.executeModelMetrics(machineInfo, modelValue);
        metrics.setModelId(modelId);

        metricsList.add(metrics);
    }

    metricsHolder.put(alias, dynMetricsList);
}

问题陈述：-

有什么方法可以让模型执行多线程，然后将结果存储到metricsList 对象中？

Answer 1

如果您的代码中没有数据竞争条件（多线程访问未正确同步的相同数据），那么您可以使用ExecutorService 形式的线程池来运行Callable线程上的实现。您可以在Callable 中完成工作。

在主线程上，ExecutorService 返回一个Future，然后您可以在所有任务都生成后等待。

线程池的大小（这里设置为10，但您可以更改它）决定了并行运行的数量。你仍然可以执行比线程池大小更多的Callables。

    Map<String, List<ServerMetrics>> metricsHolder = new HashMap<String, List<ServerMetrics>>();
    // Size of thread pool set at 10 - can be increased but increasing it 
    // to more than the number of cores on your computer is probably not 
    // useful, as it seems like your task is CPU-bound
    ExecutorService executorService = Executors.newFixedThreadPool(10);

    for (String alias : serverNames) {
        List<ClientsModel> modelMetadata = getModelAttributes();
        List<MachineInfo> machineInfo = getMachineInfo();

        List<Future<ServerMetrics>> metricsFutureList = new ArrayList<Future<ServerMetrics>>();

        // calling model one by one sequentially
        // is there any way to make this multithreaded?
        for (ClientsModel modelList : modelMetadata) {
            final String modelValue = modelList.getModelValue();
            final String modelId = String.valueOf(modelList.getModelId());

            metricsFutureList.add(executorService.submit(new Callable<ServerMetrics>() {
                @Override
                public ServerMetrics call() throws Exception {

                    // execute my model here and storing the metrics in the list
                    ServerMetrics metrics = TestUtils.executeModelMetrics(machineInfo, modelValue);
                    metrics.setModelId(modelId);
                    return metrics;
                }
            }));

        }
        List<ServerMetrics> metricsList = new ArrayList<ServerMetrics>();
        for (Future<ServerMetrics> future : metricsFutureList) {
            metricsList.add(future.get());
        }
        metricsHolder.put(alias, dynMetricsList);
    }