在互聯(lián)網(wǎng)應用中，流量洪峰是常有的事情。在應對流量洪峰時，通用的處理模式一般有排隊、限流，這樣可以非常直接有效的保護系統(tǒng)，防止系統(tǒng)被打爆。另外，通過限流技術手段，可以讓整個系統(tǒng)的運行更加平穩(wěn)。今天要與大家分享一下限流算法和C#版本的組件。

一、令牌桶算法：

??? 令牌桶算法的基本過程如下：

假如用戶配置的平均發(fā)送速率為r，則每隔1/r秒速率將一個令牌被加入到桶中；

假設桶最多可以存發(fā)b個令牌。當桶中的令牌達到上限后，丟棄令牌。

當一個有請求到達時，首先去令牌桶獲取令牌，能夠取到，則處理這個請求

如果桶中沒有令牌，那么請求排隊或者丟棄

???? 工作過程包括3個階段：產(chǎn)生令牌、消耗令牌和判斷數(shù)據(jù)包是否通過。其中涉及到2個參數(shù)：令牌產(chǎn)生的速率和令牌桶的大小，這個過程的具體工作如下。

產(chǎn)生令牌：周期性的以固定速率向令牌桶中增加令牌，桶中的令牌不斷增多。如果桶中令牌數(shù)已到達上限，則丟棄多余令牌。

消費 令牌：業(yè)務程序根據(jù)具體業(yè)務情況消耗桶中的令牌。消費一次，令牌桶令牌減少一個。

判斷是否通過：判斷是否已有令牌桶是否存在有效令牌，當桶中的令牌數(shù)量可以滿足需求時，則繼續(xù)業(yè)務處理，否則將掛起業(yè)務，等待令牌。

???? 下面是C#的一個實現(xiàn)方式

class TokenBucketLimitingService: ILimitingService
??? {
??????? private LimitedQueue<object> limitedQueue = null;
??????? private CancellationTokenSource cancelToken;
??????? private Task task = null;
??????? private int maxTPS;
??????? private int limitSize;
??????? private object lckObj = new object();
??????? public TokenBucketLimitingService(int maxTPS, int limitSize)
??????? {
??????????? this.limitSize = limitSize;
??????????? this.maxTPS = maxTPS;

?????????? if (this.limitSize <= 0)
??????????????? this.limitSize = 100;
??????????? if(this.maxTPS <=0)
??????????????? this.maxTPS = 1;

?????????? limitedQueue = new LimitedQueue<object>(limitSize);
??????????? for (int i = 0; i < limitSize; i++)
??????????? {
??????????????? limitedQueue.Enqueue(new object());
??????????? }
??????????? cancelToken = new CancellationTokenSource();
??????????? task = Task.Factory.StartNew(new Action(TokenProcess), cancelToken.Token);
??????? }

?????? /// <summary>
??????? /// 定時消息令牌
??????? /// </summary>
??????? private void TokenProcess()
??????? {
??????????? int sleep = 1000 / maxTPS;
??????????? if (sleep == 0)
??????????????? sleep = 1;

?????????? DateTime start = DateTime.Now;
??????????? while (cancelToken.Token.IsCancellationRequested ==false)
??????????? {
??????????????? try
??????????????? {
??????????????????? lock (lckObj)
??????????????????? {
??????????????????????? limitedQueue.Enqueue(new object());
??????????????????? }
??????????????? }
??????????????? catch
??????????????? {
??????????????? }
??????????????? finally
??????????????? {
??????????????????? if (DateTime.Now - start < TimeSpan.FromMilliseconds(sleep))
??????????????????? {
??????????????????????? int newSleep = sleep - (int)(DateTime.Now - start).TotalMilliseconds;
??????????????????????? if (newSleep > 1)
??????????????????????????? Thread.Sleep(newSleep - 1); //做一下時間上的補償
??????????????????? }
??????????????????? start = DateTime.Now;
??????????????? }
??????????? }
??????? }

?????? public void Dispose()
??????? {
??????????? cancelToken.Cancel();
??????? }

?????? /// <summary>
??????? /// 請求令牌
??????? /// </summary>
??????? /// <returns>true：獲取成功，false：獲取失敗</returns>
??????? public bool Request()
??????? {
??????????? if (limitedQueue.Count <= 0)
??????????????? return false;
??????????? lock (lckObj)
??????????? {
??????????????? if (limitedQueue.Count <= 0)
??????????????????? return false;

?????????????? object data = limitedQueue.Dequeue();
??????????????? if (data == null)
??????????????????? return false;
??????????? }

?????????? return true;
??????? }
??? }

public interface ILimitingService:IDisposable
???? {
???????? /// <summary>
???????? /// 申請流量處理
???????? /// </summary>
???????? /// <returns>true：獲取成功，false：獲取失敗</returns>
???????? bool Request();
???? }

public class LimitingFactory
???? {
???????? /// <summary>
???????? /// 創(chuàng)建限流服務對象
???????? /// </summary>
???????? /// <param name="limitingType">限流模型</param>
???????? /// <param name="maxQPS">最大QPS</param>
???????? /// <param name="limitSize">最大可用票據(jù)數(shù)</param>
???????? public static ILimitingService Build(LimitingType limitingType = LimitingType.TokenBucket, int maxQPS = 100, int limitSize = 100)
???????? {
???????????? switch (limitingType)
???????????? {
???????????????? case LimitingType.TokenBucket:
???????????????? default:
???????????????????? return new TokenBucketLimitingService(maxQPS, limitSize);
???????????????? case LimitingType.LeakageBucket:
???????????????????? return new LeakageBucketLimitingService(maxQPS, limitSize);
???????????? }
???????? }
???? }

?? /// <summary>
???? /// 限流模式
???? /// </summary>
???? public enum LimitingType
???? {
???????? TokenBucket,//令牌桶模式
???????? LeakageBucket//漏桶模式
???? }

public class LimitedQueue<T> : Queue<T>
???? {
???????? private int limit = 0;
???????? public const string QueueFulled = "TTP-StreamLimiting-1001";

??????? public int Limit
???????? {
???????????? get { return limit; }
???????????? set { limit = value; }
???????? }

??????? public LimitedQueue()
???????????? : this(0)
???????? { }

??????? public LimitedQueue(int limit)
???????????? : base(limit)
???????? {
???????????? this.Limit = limit;
???????? }

??????? public new bool Enqueue(T item)
???????? {
???????????? if (limit > 0 && this.Count >= this.Limit)
???????????? {
???????????????? return false;
???????????? }
???????????? base.Enqueue(item);
???????????? return true;
???????? }
???? }

???? 調(diào)用方法：

var service = LimitingFactory.Build(LimitingType.TokenBucket, 500, 200);

while (true)
{
????? var result = service.Request();
?????? //如果返回true，說明可以進行業(yè)務處理，否則需要繼續(xù)等待
?????? if (result)
?????? {
???????????? //業(yè)務處理......
?????? }
?????? else
???????????? Thread.Sleep(1);
}

二、漏桶算法

????? 聲明一個固定容量的桶，每接受到一個請求向桶中添加一個令牌，當令牌桶達到上線后請求丟棄或等待，具體算法如下：

創(chuàng)建一個固定容量的漏桶，請求到達時向漏桶添加一個令牌

如果請求添加令牌不成功，請求丟棄或等待

另一個線程以固定的速率消費桶里的令牌

???? 工作過程也包括3個階段：產(chǎn)生令牌、消耗令牌和判斷數(shù)據(jù)包是否通過。其中涉及到2個參數(shù)：令牌自動消費的速率和令牌桶的大小，個過程的具體工作如下。

產(chǎn)生令牌：業(yè)務程序根據(jù)具體業(yè)務情況申請令牌。申請一次，令牌桶令牌加一。如果桶中令牌數(shù)已到達上限，則掛起業(yè)務后等待令牌。

消費令牌：周期性的以固定速率消費令牌桶中令牌，桶中的令牌不斷較少。

判斷是否通過：判斷是否已有令牌桶是否存在有效令牌，當桶中的令牌數(shù)量可以滿足需求時，則繼續(xù)業(yè)務處理，否則將掛起業(yè)務，等待令牌。

??? C#的一個實現(xiàn)方式：

class LeakageBucketLimitingService: ILimitingService
???? {
???????? private LimitedQueue<object> limitedQueue = null;
???????? private CancellationTokenSource cancelToken;
???????? private Task task = null;
???????? private int maxTPS;
???????? private int limitSize;
???????? private object lckObj = new object();
???????? public LeakageBucketLimitingService(int maxTPS, int limitSize)
???????? {
???????????? this.limitSize = limitSize;
???????????? this.maxTPS = maxTPS;

??????????? if (this.limitSize <= 0)
???????????????? this.limitSize = 100;
???????????? if (this.maxTPS <= 0)
???????????????? this.maxTPS = 1;

??????????? limitedQueue = new LimitedQueue<object>(limitSize);
???????????? cancelToken = new CancellationTokenSource();
???????????? task = Task.Factory.StartNew(new Action(TokenProcess), cancelToken.Token);
???????? }

??????? private void TokenProcess()
???????? {
???????????? int sleep = 1000 / maxTPS;
???????????? if (sleep == 0)
???????????????? sleep = 1;

??????????? DateTime start = DateTime.Now;
???????????? while (cancelToken.Token.IsCancellationRequested == false)
???????????? {
???????????????? try
???????????????? {

??????????????????? if (limitedQueue.Count > 0)
???????????????????? {
???????????????????????? lock (lckObj)
???????????????????????? {
???????????????????????????? if (limitedQueue.Count > 0)
???????????????????????????????? limitedQueue.Dequeue();
???????????????????????? }
???????????????????? }
???????????????? }
???????????????? catch
???????????????? {
???????????????? }
???????????????? finally
???????????????? {
???????????????????? if (DateTime.Now - start < TimeSpan.FromMilliseconds(sleep))
???????????????????? {
???????????????????????? int newSleep = sleep - (int)(DateTime.Now - start).TotalMilliseconds;
???????????????????????? if (newSleep > 1)
???????????????????????????? Thread.Sleep(newSleep - 1); //做一下時間上的補償
???????????????????? }
???????????????????? start = DateTime.Now;
???????????????? }
???????????? }
???????? }

??????? public void Dispose()
???????? {
???????????? cancelToken.Cancel();
???????? }

??????? public bool Request()
???????? {
???????????? if (limitedQueue.Count >= limitSize)
???????????????? return false;
???????????? lock (lckObj)
???????????? {
???????????????? if (limitedQueue.Count >= limitSize)
???????????????????? return false;

??????????????? return limitedQueue.Enqueue(new object());
???????????? }
???????? }
???? }

??? 調(diào)用方法：

var service = LimitingFactory.Build(LimitingType.LeakageBucket, 500, 200);

while (true)
{
????? var result = service.Request();
?????? //如果返回true，說明可以進行業(yè)務處理，否則需要繼續(xù)等待
????? if (result)
?????? {
???????????? //業(yè)務處理......
?????? }
?????? else
??????????? Thread.Sleep(1);
}

兩類限流算法雖然非常相似，但是還是有些區(qū)別的，供大家參考！

漏桶算法能夠強行限制數(shù)據(jù)的傳輸速率。在某些情況下，漏桶算法不能夠有效地使用網(wǎng)絡資源。因為漏桶的漏出速率是固定的。

令牌桶算法能夠在限制數(shù)據(jù)的平均傳輸速率的同時還允許某種程度的突發(fā)傳輸.

https://www.cnblogs.com/vveiliang/p/9049393.html

總結

以上是生活随笔為你收集整理的基于.net的分布式系统限流组件的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯，歡迎將生活随笔推薦給好友。