限流系列

开源组件 rate-limit: 限流

高可用之限流-01-入门介绍

高可用之限流-02-如何设计限流框架

高可用之限流-03-Semaphore 信号量做限流

高可用之限流-04-fixed window 固定窗口

高可用之限流-05-slide window 滑动窗口

高可用之限流-06-slide window 滑动窗口 sentinel 源码

高可用之限流-07-token bucket 令牌桶算法

高可用之限流 08-leaky bucket漏桶算法

高可用之限流 09-guava RateLimiter 入门使用简介 & 源码分析

RateLimiter 入门使用

maven 引入

<dependency>
    <groupId>com.google.guava</groupId>
    <artifactId>guava</artifactId>
    <version>18.0</version>
</dependency>

测试案例

@Test
public void limitTest() {
    RateLimiter limiter = RateLimiter.create(1);
    for(int i = 1; i < 5; i++) {
        double waitTime = limiter.acquire(i);
        System.out.println("cutTime=" + System.currentTimeMillis() + " acq:" + i + " waitTime:" + waitTime);
    }
}
  • 日志
cutTime=1592880664419 acq:1 waitTime:0.0
cutTime=1592880665420 acq:2 waitTime:0.999098
cutTime=1592880667419 acq:3 waitTime:1.99867
cutTime=1592880670419 acq:4 waitTime:2.999099

说明

首先通过RateLimiter.create(1);创建一个限流器,参数代表每秒生成的令牌数,通过limiter.acquire(i);来以阻塞的方式获取令牌,当然也可以通过tryAcquire(int permits, long timeout, TimeUnit unit)来设置等待超时时间的方式获取令牌,如果超timeout为0,则代表非阻塞,获取不到立即返回。

从输出来看,RateLimiter支持预消费,比如在acquire(5)时,等待时间是3秒,是上一个获取令牌时预消费了3个两排,固需要等待3*1秒,然后又预消费了5个令牌,以此类推

RateLimiter通过限制后面请求的等待时间,来支持一定程度的突发请求(预消费),在使用过程中需要注意这一点,具体实现原理后面再分析。

RateLimiter实现原理

Guava有两种限流模式,一种为稳定模式(SmoothBursty:令牌生成速度恒定),一种为渐进模式(SmoothWarmingUp:令牌生成速度缓慢提升直到维持在一个稳定值) 两种模式实现思路类似,主要区别在等待时间的计算上,本篇重点介绍SmoothBursty

RateLimiter的创建

通过调用RateLimiter的create接口来创建实例,实际是调用的SmoothBuisty稳定模式创建的实例。

public static RateLimiter create(double permitsPerSecond) {
  return create(permitsPerSecond, SleepingStopwatch.createFromSystemTimer());
}

static RateLimiter create(double permitsPerSecond, SleepingStopwatch stopwatch) {
  RateLimiter rateLimiter = new SmoothBursty(stopwatch, 1.0 /* maxBurstSeconds */);
  rateLimiter.setRate(permitsPerSecond);
  return rateLimiter;
}

SmoothBursty中的两个构造参数含义:

SleepingStopwatch:guava中的一个时钟类实例,会通过这个来计算时间及令牌

maxBurstSeconds:官方解释,在ReteLimiter未使用时,最多保存几秒的令牌,默认是1

在解析SmoothBursty原理前,重点解释下SmoothBursty中几个属性的含义

/**
 * The work (permits) of how many seconds can be saved up if this RateLimiter is unused?
 * 在RateLimiter未使用时,最多存储几秒的令牌
 * */
 final double maxBurstSeconds;
 

/**
 * The currently stored permits.
 * 当前存储令牌数
 */
double storedPermits;

/**
 * The maximum number of stored permits.
 * 最大存储令牌数 = maxBurstSeconds * stableIntervalMicros(见下文)
 */
double maxPermits;

/**
 * The interval between two unit requests, at our stable rate. E.g., a stable rate of 5 permits
 * per second has a stable interval of 200ms.
 * 添加令牌时间间隔 = SECONDS.toMicros(1L) / permitsPerSecond;(1秒/每秒的令牌数)
 */
double stableIntervalMicros;

/**
 * The time when the next request (no matter its size) will be granted. After granting a request,
 * this is pushed further in the future. Large requests push this further than small requests.
 * 下一次请求可以获取令牌的起始时间
 * 由于RateLimiter允许预消费,上次请求预消费令牌后
 * 下次请求需要等待相应的时间到nextFreeTicketMicros时刻才可以获取令牌
 */
private long nextFreeTicketMicros = 0L; // could be either in the past or future

核心函数

setRate()

通过这个接口设置令牌通每秒生成令牌的数量,内部时间通过调用SmoothRateLimiter的doSetRate来实现

public final void setRate(double permitsPerSecond) {
  checkArgument(
      permitsPerSecond > 0.0 && !Double.isNaN(permitsPerSecond), "rate must be positive");
  synchronized (mutex()) {
    doSetRate(permitsPerSecond, stopwatch.readMicros());
  }
}

doSetRate()

这里先通过调用resync生成令牌以及更新下一期令牌生成时间,然后更新stableIntervalMicros,最后又调用了SmoothBursty的doSetRate

@Override
final void doSetRate(double permitsPerSecond, long nowMicros) {
  resync(nowMicros);
  double stableIntervalMicros = SECONDS.toMicros(1L) / permitsPerSecond;
  this.stableIntervalMicros = stableIntervalMicros;
  doSetRate(permitsPerSecond, stableIntervalMicros);
}

resync()

/**
 * Updates {@code storedPermits} and {@code nextFreeTicketMicros} based on the current time.
 * 基于当前时间,更新下一次请求令牌的时间,以及当前存储的令牌(可以理解为生成令牌)
 */
void resync(long nowMicros) {
    // if nextFreeTicket is in the past, resync to now
    if (nowMicros > nextFreeTicketMicros) {
      double newPermits = (nowMicros - nextFreeTicketMicros) / coolDownIntervalMicros();
      storedPermits = min(maxPermits, storedPermits + newPermits);
      nextFreeTicketMicros = nowMicros;
    }
}

根据令牌桶算法,桶中的令牌是持续生成存放的,有请求时需要先从桶中拿到令牌才能开始执行,谁来持续生成令牌存放呢?

一种解法是,开启一个定时任务,由定时任务持续生成令牌。这样的问题在于会极大的消耗系统资源,如,某接口需要分别对每个用户做访问频率限制,假设系统中存在6W用户,则至多需要开启6W个定时任务来维持每个桶中的令牌数,这样的开销是巨大的。

另一种解法则是延迟计算,如上resync函数。该函数会在每次获取令牌之前调用,其实现思路为,若当前时间晚于nextFreeTicketMicros,则计算该段时间内可以生成多少令牌,将生成的令牌加入令牌桶中并更新数据。这样一来,只需要在获取令牌时计算一次即可。

SmoothBursty 的 doSetRate

桶中可存放的最大令牌数由maxBurstSeconds计算而来,其含义为最大存储maxBurstSeconds秒生成的令牌。 该参数的作用在于,可以更为灵活地控制流量。如,某些接口限制为300次/20秒,某些接口限制为50次/45秒等。也就是流量不局限于qps

作者:人在码途 链接:https://www.jianshu.com/p/5d4fe4b2a726 来源:简书 著作权归作者所有。商业转载请联系作者获得授权,非商业转载请注明出处。

@Override
void doSetRate(double permitsPerSecond, double stableIntervalMicros) {
  double oldMaxPermits = this.maxPermits;
  maxPermits = maxBurstSeconds * permitsPerSecond;
  if (oldMaxPermits == Double.POSITIVE_INFINITY) {
    // if we don't special-case this, we would get storedPermits == NaN, below
    // Double.POSITIVE_INFINITY 代表无穷啊
    storedPermits = maxPermits;
  } else {
    storedPermits =
        (oldMaxPermits == 0.0)
            ? 0.0 // initial state
            : storedPermits * maxPermits / oldMaxPermits;
  }
}

RateLimiter 几个常用接口分析

在了解以上概念后,就非常容易理解 RateLimiter 暴露出来的接口

@CanIgnoreReturnValue
public double acquire() {
  return acquire(1);
}

/**
* 获取令牌,返回阻塞的时间
**/
@CanIgnoreReturnValue
public double acquire(int permits) {
  long microsToWait = reserve(permits);
  stopwatch.sleepMicrosUninterruptibly(microsToWait);
  return 1.0 * microsToWait / SECONDS.toMicros(1L);
}

final long reserve(int permits) {
  checkPermits(permits);
  synchronized (mutex()) {
    return reserveAndGetWaitLength(permits, stopwatch.readMicros());
  }
}

acquire函数主要用于获取permits个令牌,并计算需要等待多长时间,进而挂起等待,并将该值返回,主要通过reserve返回需要等待的时间,reserve中通过调用reserveAndGetWaitLength获取等待时间

/**
 * Reserves next ticket and returns the wait time that the caller must wait for.
 *
 * @return the required wait time, never negative
 */
final long reserveAndGetWaitLength(int permits, long nowMicros) {
  long momentAvailable = reserveEarliestAvailable(permits, nowMicros);
  return max(momentAvailable - nowMicros, 0);
}

最后调用了 reserveEarliestAvailable

@Override
final long reserveEarliestAvailable(int requiredPermits, long nowMicros) {
  resync(nowMicros);
  long returnValue = nextFreeTicketMicros;
  double storedPermitsToSpend = min(requiredPermits, this.storedPermits);
  double freshPermits = requiredPermits - storedPermitsToSpend;
  long waitMicros =
      storedPermitsToWaitTime(this.storedPermits, storedPermitsToSpend)
          + (long) (freshPermits * stableIntervalMicros);

  this.nextFreeTicketMicros = LongMath.saturatedAdd(nextFreeTicketMicros, waitMicros);
  this.storedPermits -= storedPermitsToSpend;
  return returnValue;
}

首先通过resync生成令牌以及同步nextFreeTicketMicros时间戳,freshPermits从令牌桶中获取令牌后还需要的令牌数量,通过storedPermitsToWaitTime计算出获取freshPermits还需要等待的时间,在稳定模式中,这里就是(long) (freshPermits * stableIntervalMicros) ,然后更新nextFreeTicketMicros以及storedPermits,这次获取令牌需要的等待到的时间点,reserveAndGetWaitLength返回需要等待的时间间隔。

reserveEarliestAvailable可以看出RateLimiter的预消费原理,以及获取令牌的等待时间时间原理(可以解释示例结果),再获取令牌不足时,并没有等待到令牌全部生成,而是更新了下次获取令牌时的nextFreeTicketMicros,从而影响的是下次获取令牌的等待时间。

reserve这里返回等待时间后,acquire通过调用stopwatch.sleepMicrosUninterruptibly(microsToWait);进行sleep操作,这里不同于Thread.sleep(), 这个函数的sleep是uninterruptibly的,内部实现:

public static void sleepUninterruptibly(long sleepFor, TimeUnit unit) {
    //sleep 阻塞线程 内部通过Thread.sleep()
  boolean interrupted = false;
  try {
    long remainingNanos = unit.toNanos(sleepFor);
    long end = System.nanoTime() + remainingNanos;
    while (true) {
      try {
        // TimeUnit.sleep() treats negative timeouts just like zero.
        NANOSECONDS.sleep(remainingNanos);
        return;
      } catch (InterruptedException e) {
        interrupted = true;
        remainingNanos = end - System.nanoTime();
        //如果被interrupt可以继续,更新sleep时间,循环继续sleep
      }
    }
  } finally {
    if (interrupted) {
      Thread.currentThread().interrupt();
      //如果被打断过,sleep过后再真正中断线程
    }
  }
}

sleep之后,acquire返回sleep的时间,阻塞结束,获取到令牌。

public boolean tryAcquire(int permits) {
  return tryAcquire(permits, 0, MICROSECONDS);
}

public boolean tryAcquire() {
  return tryAcquire(1, 0, MICROSECONDS);
}

public boolean tryAcquire(int permits, long timeout, TimeUnit unit) {
  long timeoutMicros = max(unit.toMicros(timeout), 0);
  checkPermits(permits);
  long microsToWait;
  synchronized (mutex()) {
    long nowMicros = stopwatch.readMicros();
    if (!canAcquire(nowMicros, timeoutMicros)) {
      return false;
    } else {
      microsToWait = reserveAndGetWaitLength(permits, nowMicros);
    }
  }
  stopwatch.sleepMicrosUninterruptibly(microsToWait);
  return true;
}

private boolean canAcquire(long nowMicros, long timeoutMicros) {
  return queryEarliestAvailable(nowMicros) - timeoutMicros <= nowMicros;
}

@Override
final long queryEarliestAvailable(long nowMicros) {
  return nextFreeTicketMicros;
}

tryAcquire函数可以尝试在timeout时间内获取令牌,如果可以则挂起等待相应时间并返回true,否则立即返回false

canAcquire用于判断timeout时间内是否可以获取令牌,通过判断当前时间+超时时间是否大于nextFreeTicketMicros 来决定是否能够拿到足够的令牌数,如果可以获取到,则过程同acquire,线程sleep等待,如果通过canAcquire在此超时时间内不能回去到令牌,则可以快速返回,不需要等待timeout后才知道能否获取到令牌。

到此,Guava RateLimiter稳定模式的实现原理基本已经清楚,如发现文中错误的地方,劳烦指正!

参考资料

限流限速 RateLimiter

Guava RateLimiter