限流的不完全总结

偏执的太偏执、 2023-09-25 12:49 102阅读 0赞

限流可以在网关这一层做，比如Nginx、Openresty、kong、zuul、Spring Cloud Gateway等；也可以在应用层通过Aop这种方式去做限流。

一些限流组件，包括Redisson的令牌桶限流，Guava的RateLimiter。Hystrix是客户端限流（参考[彗安金科Java面试（部分）][Java] 或直接参考[熔断器Hystrix][Hystrix] 或[JavaGuide—Java面试指北—高性能：负载均衡的常见算法有哪些？][JavaGuide_Java]）。

## 网关层限流 ##

![在这里插入图片描述][3c52f2c8f9d34a71a7b4c433c1fb665d.png]1

网关层限流：gateway的限流基于Redis+Lua脚本1 2 :

[Redisson分布式限流器RRateLimiter原理解析][Redisson_RRateLimiter]

## 单机限流：可Hystrix用线程池或Semaphore限流： ##

例1：[深入 Hystrix 线程池隔离与接口限流][Hystrix 1]

实例2：  
RedisHystrixCommand.java :

public class RedisHystrixCommand<T> extends HystrixCommand<T> {
        
    
      private static final Logger logger = new Logger("Hystrix");
    
      private Callable<T> callable;
      private T defaultResult;
      private String opName;
    
      /**
       * constructor.
       */
      public RedisHystrixCommand(String keyName, Callable<T> callable, T defaultResult,
          String opName) {
        
        // default timeout=1s, fallback when 50% or 20 requests in 10 seconds fail
        super(Setter
            .withGroupKey(HystrixCommandGroupKey.Factory.asKey(keyName))
            .andCommandKey(HystrixCommandKey.Factory.asKey(keyName))
            .andCommandPropertiesDefaults(HystrixCommandProperties.Setter()
                .withExecutionIsolationStrategy(ExecutionIsolationStrategy.SEMAPHORE)
                .withCircuitBreakerEnabled(true)));
        this.callable = callable;
        this.defaultResult = defaultResult;
        this.opName = opName;
      }
    
      @Override
      protected T run() throws Exception {
        
        return callable.call();
      }
    
      @Override
      protected T getFallback() {
        
        logger.error("RedisCache {} fallback since executionTimeout, return {}",
            opName, defaultResult);
        return defaultResult;
      }
    }

RedisHelper.java:

public Long sadd(String key, String member) {
        
      return new RedisHystrixCommand<>(getCommandKey(), () ->
          getRedisClient().handle(redis -> redis.sadd(key, member))
              .orElse(0L), 0L, "sadd").execute();
    }

--------------------

1.  [我司用了 6 年的 Redis 分布式限流器，可以说是非常厉害了！][6 _ Redis]；  
    [5 种限流算法，7 种限流方式，挡住突发流量][5 _7]  
    其中后文靠后的部分提到了用Redis+lua脚本来实现两种限流算法：一是用INCR 和 EXPIRE 命令来实现固定窗口限流，二是用ZSET（score为当前时间）实现滑动窗口限流。 ↩︎ ↩︎
2.  [Spring Cloud Gateway 之限流篇][Spring Cloud Gateway] ↩︎

[Java]: https://blog.csdn.net/qq_23204557/article/details/115154884?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522167510332016800213031434%2522%252C%2522scm%2522%253A%252220140713.130102334.pc%255Fblog.%2522%257D&request_id=167510332016800213031434&biz_id=0&utm_medium=distribute.pc_search_result.none-task-blog-2~blog~first_rank_ecpm_v1~rank_v31_ecpm-1-115154884-null-null.blog_rank_default&utm_term=feign&spm=1018.2226.3001.4450
[Hystrix]: http://www.ityouknow.com/springcloud/2017/05/16/springcloud-hystrix.html
[JavaGuide_Java]: https://www.yuque.com/snailclimb/mf2z3k/dxxpey
[3c52f2c8f9d34a71a7b4c433c1fb665d.png]: https://img-blog.csdnimg.cn/3c52f2c8f9d34a71a7b4c433c1fb665d.png
[Redisson_RRateLimiter]: https://github.com/oneone1995/blog/issues/13
[Hystrix 1]: https://www.jianshu.com/p/847613a84e5e
[6 _ Redis]: https://mp.weixin.qq.com/s/kyFAWH3mVNJvurQDt4vchA
[5 _7]: https://www.wdbyte.com/java/rate-limiter/
[Spring Cloud Gateway]: https://mp.weixin.qq.com/s?__biz=MzI4NjE4NTUwNQ==&mid=2247494915&idx=5&sn=e30db209b5af1b5e760a3a4f2f2b7e12&chksm=ebe26c4ddc95e55b43f93fa2dad4209ff4796bff74d6281ca740f84d99a8e2a059010abc48cd&scene=27