如何在多线程应用程序中正确使用随机类

Question

我需要使用随机类在公共静态函数内的多线程应用程序中生成随机数。我怎样才能实现它。目前，下面的功能运行良好，但与随机类相比并不是很快。所以我需要修改下面的函数并使其与随机类一起工作，而该类正在发生数千个并发调用。如果我使用随机它对我认为的每个调用都使用相同的种子，并且随机化非常糟糕。 我现在的班级

using System;
using System.Collections.Generic;
using System.Linq;
using System.Web;
using System.Security.Cryptography;

public static class GenerateRandomValue
{
    static RNGCryptoServiceProvider Gen = new RNGCryptoServiceProvider();

    public static int GenerateRandomValueDefault(int irRandValRange)//default min val 1
    {
        if (irRandValRange == 0)
            irRandValRange = 1;
        byte[] randomNumber = new byte[4]; // 4 bytes per Int32
        Gen.GetBytes(randomNumber);
        return Math.Abs(BitConverter.ToInt32(randomNumber, 0) % irRandValRange) + 1;
    }

    public static int GenerateRandomValueMin(int irRandValRange, int irMinValue)
    {
        byte[] randomNumber = new byte[4]; // 4 bytes per Int32
        Gen.GetBytes(randomNumber);
        return BitConverter.ToInt32(randomNumber, 0) % irRandValRange + irMinValue;
    }
}

另一个看起来不错且线程安全的函数

using System;
using System.Collections.Generic;
using System.Linq;
using System.Web;
using System.Threading;

public static class GenerateRandomValue
{
    private static Random seedGenerator = new Random();

    private static ThreadLocal<Random> random = new ThreadLocal<Random>(SeededRandomFactory);

    private static Random SeededRandomFactory()
    {
        lock(seedGenerator)
            return new Random(seedGenerator.Next());
    }

    public static int GenerateRandomValueMin(int irRandValRange, int irMinValue)
    {
        return random.Value.Next(irMinValue, irRandValRange);
    }
}

Answer 1

您需要的是一种更好的方式来在您的 ASP.NET 应用程序中开始播种，使用下面的方法，Random 的质量应该没问题。

public static int GenerateRandomValueDefault(int irRandValRange)//default min val 1
{
    return GenerateRandomValueMin(irRandValRange, 1);
}

public static int GenerateRandomValueMin(int irRandValRange, int irMinValue)
{
    Random rand = GetRandom();
    return rand.GetNext(irMinValue,irRandValRange)
}

//This is a global random number generator, it is only used to provide the seed for the local RNG's.
private static Random GlobalRandom = new Random();

private static Random GetRandom()
{
    if (HttpContext.Current.Session["RNG"] == null)
    {
        //This lock is only hit the very first time the users Session state is used, every time after that it should use the cached local copy without taking the lock.
        lock(GlobalRandom)
        {
            //We use the Global RNG for seed instead of the default because it uses time to seed by default, and if two people get a new Random() within the same time-slice they will have the same seed. This prevents that from happening.
            HttpContext.Current.Session["RNG"] = new Random(GlobalRandom.Next());
        }
    }
    //Return the cached/new RNG.
    return (Random)HttpContext.Current.Session["RNG"];
}

您有一个全局 RNG 实例确实会锁定，但是只有在生成新的会话状态时才会发生这种情况，之后会话仅使用它的本地副本。您将在运行时获得非常好的性能，因为它会从全局存储中生成一个数字，因此每个人的第一页加载会产生轻微的负载。

您可以对其进行修改以满足您的需要，但它会为您提供总体思路，但也可以为您提供总体思路。

---

根据 Henk Holterman 的建议，这是一种无锁解决方案，它可能更快且不使用 HttpState。

private static int SeedCounter = 0;
private readonly object SeedInitLock = new Object();

private static Random GetRandom()
{
    //Do init the first time this function is ever called.
    if(SeedCounter == -1)
    {
        //The first time the function is called everyone will try to update SeedCounter, but only the first 
        //thread to complete it will be the value everyone uses.
        Random initRNG = new Random();
        Interlocked.CompareExchange(ref SeedCounter, initRNG.Next(), -1);

    }
    else if (SeedCounter < 0)
    {
        //Because Interlocked.Increment wraps the value to int.MinValue and Random(int) will take the absolute
        //value of the seed, we skip all of the negitive numbers and go to 0.
        Interlocked.CompareExchange(ref SeedCounter, 0, int.MinValue);
    }

    int tempSeed = Interlocked.Increment(ref SeedCounter);
    if (tempSeed < 0)
    {
        //If tempSeed is negative we hit a edge case where SeedCounter wrapped around. We just call the function
        //again so we do not reuse a seed that was just used.
        return GetRandom();
    }

    return new Random(tempSeed);
}

Answer 2

您没有指定除速度以外的任何限制，所以我认为Random 应该这样做。

// Field in the class
Random rand = new Random();

// Inside a method:
int randomValue = rand.Next(); // Random positive integer returned
int randomValue = rand.Next(max); // Random under max
int randomValue = rand.Next(min, max); // Random in range

Answer 3

我建议System.Web.Security.Membership.GeneratePassword() 方法。

string generated = System.Web.Security.Membership.GeneratePassword(
                   10, // maximum length
                   3)  // number of non-ASCII characters.

Answer 4

使用具有良好种子的 Random 被认为是“性能方面”更好。但是你需要性能吗？

查看此博客的指标：http://blogs.msdn.com/b/pfxteam/archive/2009/02/19/9434171.aspx 最后有一个关于与提供者播种的提示。

不过，还是要加点盐。您必须重做测试才能确定。