C# 并行编程 之 限制资源的并发访问 使用SemaphoreSlim

概要

当多个任务或线程并行运行时，难以避免的对某些有限的资源进行并发的访问。可以考虑使用信号量来进行这方面的控制（System.Threading.Semaphore）是表示一个Windows内核的信号量对象。如果预计等待的时间较短，可以考虑使用SemaphoreSlim，它则带来的开销更小。

.NetFrameWork中的信号量通过跟踪进入和离开的任务或线程来协调对资源的访问。信号量需要知道资源的最大数量，当一个任务进入时，资源计数器会被减1，当计数器为0时，如果有任务访问资源，它会被阻塞，直到有任务离开为止。

示例程序： 10个任务并行访问3个资源

using System;
using System.Text;
using System.Threading;
using System.Threading.Tasks;
using System.Diagnostics;

namespace Sample5_8_semaphoreslim
{
    class Program
    {
        private static int _TaskNum = 10;
        private static Task[] _Tasks;
        private const int MAX_RESOURCE = 3;
        private const int RUN_LOOP = 10;

        private static SemaphoreSlim m_Semaphore;

        private static void Work1(int TaskID)
        {
            int i = 0;
            var sw = Stopwatch.StartNew();
            var rnd = new Random();

            while (i < RUN_LOOP)
            {
                Thread.Sleep(rnd.Next(200, 500));

                Console.WriteLine("TASK " + TaskID + " REQUESTing {");
                m_Semaphore.Wait();

                try
                {
                    Console.WriteLine("TASK " + TaskID + " WOrking ... ..." + i);
                    sw.Restart();
                    Thread.Sleep(rnd.Next(200, 500));
                }
                finally
                {
                    Console.WriteLine("TASK " + TaskID + " REQUESTing }");
                    m_Semaphore.Release();
                    i++;
                }
            }
        }

        static void Main(string[] args)
        {
            _Tasks = new Task[_TaskNum];
            m_Semaphore = new SemaphoreSlim(MAX_RESOURCE);
            int i = 0;

            for (i = 0; i < _TaskNum; i++)
            {
                _Tasks[i] = Task.Factory.StartNew((num) =>
                {
                    var taskid = (int)num;
                    Work1(taskid);
                }, i);
            }

            var finalTask = Task.Factory.ContinueWhenAll(_Tasks, (tasks) =>
            {
                Task.WaitAll(_Tasks);
                Console.WriteLine("==========================================================");
                Console.WriteLine("All Phase is completed");
                Console.WriteLine("==========================================================");
            });

            try
            {
                finalTask.Wait();
            }
            catch (AggregateException aex)
            {
                Console.WriteLine("Task failed And Canceled" + aex.ToString());
            }
            finally
            {
                m_Semaphore.Dispose();
            }
            Console.ReadLine();
        }
    }
}

使用超时和取消

信号量当然不可能永久的阻塞在那里。信号量也提供了超时处理机制。方法是在Wait函数中传入一个超时等待时间 - Wait(int TIMEOUT)。当Wait返回值为false时表明它超时了。如果传入了 -1，则表示无限期的等待。

程序示例：注意其中的m_Semaphore.Release();已经被注释掉了，任务会等待1秒钟然后超时。

using System;
using System.Text;
using System.Threading;
using System.Threading.Tasks;
using System.Diagnostics;

namespace Sample5_8_semaphoreslim
{
    class Program
    {
        private static int _TaskNum = 10;
        private static Task[] _Tasks;
        private const int MAX_RESOURCE = 3;
        private const int RUN_LOOP = 10;

        private static SemaphoreSlim m_Semaphore;

        private static void Work1(int TaskID)
        {
            int i = 0;
            var sw = Stopwatch.StartNew();
            var rnd = new Random();

            while (i < RUN_LOOP)
            {
                Thread.Sleep(rnd.Next(200, 500));

                Console.WriteLine("TASK " + TaskID + " REQUESTing {");
                if (!m_Semaphore.Wait(1000))
                {
                    Console.WriteLine("TASK " + TaskID + " TIMEOUT!!!");
                    return;
                }

                try
                {
                    Console.WriteLine("TASK " + TaskID + " WOrking ... ..." + i);
                    sw.Restart();
                    Thread.Sleep(rnd.Next(2000, 5000));
                }
                finally
                {
                    Console.WriteLine("TASK " + TaskID + " REQUESTing }");
                    //m_Semaphore.Release();
                    i++;
                }
            }
        }


        static void Main(string[] args)
        {
            _Tasks = new Task[_TaskNum];
            m_Semaphore = new SemaphoreSlim(MAX_RESOURCE);
            int i = 0;

            for (i = 0; i < _TaskNum; i++)
            {
                _Tasks[i] = Task.Factory.StartNew((num) =>
                {
                    var taskid = (int)num;
                    Work1(taskid);
                }, i);
            }

            var finalTask = Task.Factory.ContinueWhenAll(_Tasks, (tasks) =>
            {
                Task.WaitAll(_Tasks);
                Console.WriteLine("==========================================================");
                Console.WriteLine("All Phase is completed");
                Console.WriteLine("==========================================================");
            });

            try
            {
                finalTask.Wait();
            }
            catch (AggregateException aex)
            {
                Console.WriteLine("Task failed And Canceled" + aex.ToString());
            }
            finally
            {
                m_Semaphore.Dispose();
            }
            Console.ReadLine();
        }
    }
}

跨进程或AppDomain的同步

如果需要有跨进程或AppDomain的同步时，可以考虑使用Semaphore。Semaphore是取得的Windows 内核的信号量，所以在整个系统中是有效的。
它主要的接口时 Release和WaitOne，使用的方式和SemaphoreSlim是一致的。