PHP后端限流算法比较-源码库

PHP后端限流算法比较：从简单计数到漏桶与令牌桶实战

上周在优化API网关时，我又遇到了那个熟悉的问题：某个热门接口突然被刷，导致整个服务响应变慢。经过一番折腾，我重新梳理了PHP中几种常见的限流方案，今天就把这些实战经验分享给大家，特别是三种核心算法的实现对比。

1. 为什么我们需要限流？

记得第一次做电商促销活动时，由于没有做限流，瞬间涌入的请求直接把数据库打挂了。从那以后我明白了：限流不只是防止恶意请求，更是保障系统稳定性的重要手段。它能：

避免单个接口耗尽服务器资源
防止爬虫过度抓取
保证核心业务的稳定性
实现服务的优雅降级

2. 固定窗口计数器：最简单的开始

这是我最早接触的限流方式，思路很简单：在固定时间窗口内计数，超过阈值就拒绝。


class FixedWindowLimiter
{
    private $redis;
    private $key;
    private $maxRequests;
    private $windowSize;
    
    public function __construct($redis, $key, $maxRequests = 100, $windowSize = 60)
    {
        $this->redis = $redis;
        $this->key = $key;
        $this->maxRequests = $maxRequests;
        $this->windowSize = $windowSize;
    }
    
    public function isAllowed()
    {
        $current = $this->redis->get($this->key);
        
        if ($current === false || $current < $this->maxRequests) {
            $this->redis->multi();
            $this->redis->incr($this->key);
            $this->redis->expire($this->key, $this->windowSize);
            $this->redis->exec();
            return true;
        }
        
        return false;
    }
}

踩坑提醒：这种算法在窗口切换时会有临界问题。比如限制每分钟100次，可能在59秒和61秒分别请求100次，实际上2秒内请求了200次。

3. 滑动窗口：解决临界问题

为了解决固定窗口的缺陷，我采用了滑动窗口算法，记录更细粒度的时间片：


class SlidingWindowLimiter
{
    private $redis;
    private $key;
    private $maxRequests;
    private $windowSize;
    private $precision;
    
    public function __construct($redis, $key, $maxRequests = 100, $windowSize = 60, $precision = 10)
    {
        $this->redis = $redis;
        $this->key = $key;
        $this->maxRequests = $maxRequests;
        $this->windowSize = $windowSize;
        $this->precision = $precision; // 时间片精度（秒）
    }
    
    public function isAllowed()
    {
        $now = time();
        $windowStart = $now - $this->windowSize;
        
        // 移除过期时间片
        $this->redis->zRemRangeByScore($this->key, 0, $windowStart);
        
        // 获取当前窗口内请求数
        $current = $this->redis->zCard($this->key);
        
        if ($current < $this->maxRequests) {
            $this->redis->zAdd($this->key, $now, uniqid());
            $this->redis->expire($this->key, $this->windowSize * 2);
            return true;
        }
        
        return false;
    }
}

4. 漏桶算法：平滑流量输出

在处理消息队列消费时，我发现漏桶算法特别适合控制处理速率：


class LeakyBucketLimiter
{
    private $redis;
    private $key;
    private $capacity;    // 桶容量
    private $leakRate;    // 漏水速率（个/秒）
    
    public function __construct($redis, $key, $capacity = 100, $leakRate = 10)
    {
        $this->redis = $redis;
        $this->key = $key;
        $this->capacity = $capacity;
        $this->leakRate = $leakRate;
    }
    
    public function isAllowed()
    {
        $now = microtime(true);
        $data = $this->redis->hGetAll($this->key);
        
        if (empty($data)) {
            $this->redis->hMSet($this->key, [
                'water' => 1,
                'last_time' => $now
            ]);
            $this->redis->expire($this->key, 3600);
            return true;
        }
        
        $water = (float)$data['water'];
        $lastTime = (float)$data['last_time'];
        
        // 计算漏水量
        $leakAmount = ($now - $lastTime) * $this->leakRate;
        $water = max(0, $water - $leakAmount);
        
        if ($water + 1 <= $this->capacity) {
            $this->redis->hMSet($this->key, [
                'water' => $water + 1,
                'last_time' => $now
            ]);
            return true;
        }
        
        return false;
    }
}

5. 令牌桶算法：应对突发流量

在需要应对突发流量的场景下，令牌桶是我的首选：


class TokenBucketLimiter
{
    private $redis;
    private $key;
    private $capacity;    // 桶容量
    private $fillRate;    // 令牌添加速率（个/秒）
    
    public function __construct($redis, $key, $capacity = 100, $fillRate = 10)
    {
        $this->redis = $redis;
        $this->key = $key;
        $this->capacity = $capacity;
        $this->fillRate = $fillRate;
    }
    
    public function isAllowed($tokens = 1)
    {
        $now = microtime(true);
        $data = $this->redis->hGetAll($this->key);
        
        if (empty($data)) {
            $this->redis->hMSet($this->key, [
                'tokens' => $this->capacity - $tokens,
                'last_time' => $now
            ]);
            $this->redis->expire($this->key, 3600);
            return true;
        }
        
        $tokens = (float)$data['tokens'];
        $lastTime = (float)$data['last_time'];
        
        // 计算新增令牌
        $newTokens = ($now - $lastTime) * $this->fillRate;
        $tokens = min($this->capacity, $tokens + $newTokens);
        
        if ($tokens >= 1) {
            $this->redis->hMSet($this->key, [
                'tokens' => $tokens - 1,
                'last_time' => $now
            ]);
            return true;
        }
        
        return false;
    }
}