14.4 限流算法实现
面试重要程度：⭐⭐⭐⭐⭐
常见提问方式： "设计API限流系统" "令牌桶vs漏桶算法" "分布式限流实现"
预计阅读时间：40分钟
🎯 限流算法基础
常见限流算法
限流是保护系统稳定性的重要手段，常见算法包括：
- 固定窗口计数器：简单但有突刺问题
- 滑动窗口计数器：平滑限流，内存开销大
- 令牌桶算法：允许突发流量，平滑限流
- 漏桶算法：强制限制输出速率
🪙 令牌桶算法
令牌桶实现
/**
* 令牌桶限流算法
*/
public class TokenBucketRateLimiter {
private final int capacity; // 桶容量
private final double refillRate; // 令牌补充速率（个/秒）
private final Map<String, TokenBucket> buckets;
public TokenBucketRateLimiter(int capacity, double refillRate) {
this.capacity = capacity;
this.refillRate = refillRate;
this.buckets = new ConcurrentHashMap<>();
}
/**
* 令牌桶
*/
static class TokenBucket {
private volatile double tokens;
private volatile long lastRefillTime;
private final int capacity;
private final double refillRate;
public TokenBucket(int capacity, double refillRate) {
this.capacity = capacity;
this.refillRate = refillRate;
this.tokens = capacity;
this.lastRefillTime = System.currentTimeMillis();
}
public synchronized boolean tryAcquire(int tokensRequested) {
refill();
if (tokens >= tokensRequested) {
tokens -= tokensRequested;
return true;
}
return false;
}
private void refill() {
long currentTime = System.currentTimeMillis();
double timePassed = (currentTime - lastRefillTime) / 1000.0;
if (timePassed > 0) {
double tokensToAdd = timePassed * refillRate;
tokens = Math.min(capacity, tokens + tokensToAdd);
lastRefillTime = currentTime;
}
}
public double getAvailableTokens() {
refill();
return tokens;
}
}
/**
* 尝试获取许可
*/
public boolean tryAcquire(String key) {
return tryAcquire(key, 1);
}
/**
* 尝试获取指定数量的许可
*/
public boolean tryAcquire(String key, int permits) {
TokenBucket bucket = buckets.computeIfAbsent(key,
k -> new TokenBucket(capacity, refillRate));
return bucket.tryAcquire(permits);
}
}
🌐 分布式限流
基于Redis的分布式限流
/**
* 基于Redis的分布式限流器
*/
@Component
public class DistributedRateLimiter {
private final RedisTemplate<String, Object> redisTemplate;
private final String keyPrefix = "rate_limiter:";
public DistributedRateLimiter(RedisTemplate<String, Object> redisTemplate) {
this.redisTemplate = redisTemplate;
}
/**
* 固定窗口限流
*/
public boolean tryAcquireFixedWindow(String key, int maxRequests, long windowSizeInSeconds) {
String redisKey = keyPrefix + "fixed:" + key;
long currentWindow = System.currentTimeMillis() / (windowSizeInSeconds * 1000);
String windowKey = redisKey + ":" + currentWindow;
// 使用Lua脚本保证原子性
String luaScript =
"local current = redis.call('GET', KEYS[1]) " +
"if current == false then " +
" redis.call('SET', KEYS[1], 1) " +
" redis.call('EXPIRE', KEYS[1], ARGV[2]) " +
" return 1 " +
"else " +
" if tonumber(current) < tonumber(ARGV[1]) then " +
" return redis.call('INCR', KEYS[1]) " +
" else " +
" return -1 " +
" end " +
"end";
DefaultRedisScript<Long> script = new DefaultRedisScript<>(luaScript, Long.class);
Long result = redisTemplate.execute(script,
Collections.singletonList(windowKey),
maxRequests, windowSizeInSeconds);
return result != null && result != -1;
}
/**
* 令牌桶限流
*/
public boolean tryAcquireTokenBucket(String key, int capacity, double refillRate) {
String redisKey = keyPrefix + "token:" + key;
long currentTime = System.currentTimeMillis();
String luaScript =
"local bucket = redis.call('HMGET', KEYS[1], 'tokens', 'lastRefill') " +
"local tokens = tonumber(bucket[1]) or tonumber(ARGV[1]) " +
"local lastRefill = tonumber(bucket[2]) or tonumber(ARGV[4]) " +
"local timePassed = (tonumber(ARGV[4]) - lastRefill) / 1000.0 " +
"if timePassed > 0 then " +
" tokens = math.min(tonumber(ARGV[1]), tokens + timePassed * tonumber(ARGV[2])) " +
"end " +
"if tokens >= tonumber(ARGV[3]) then " +
" tokens = tokens - tonumber(ARGV[3]) " +
" redis.call('HMSET', KEYS[1], 'tokens', tokens, 'lastRefill', ARGV[4]) " +
" redis.call('EXPIRE', KEYS[1], 3600) " +
" return 1 " +
"else " +
" redis.call('HMSET', KEYS[1], 'tokens', tokens, 'lastRefill', ARGV[4]) " +
" redis.call('EXPIRE', KEYS[1], 3600) " +
" return 0 " +
"end";
DefaultRedisScript<Long> script = new DefaultRedisScript<>(luaScript, Long.class);
Long result = redisTemplate.execute(script,
Collections.singletonList(redisKey),
capacity, refillRate, 1, currentTime);
return result != null && result == 1;
}
}
🎯 限流注解和AOP
限流注解
/**
* 限流注解
*/
@Target({ElementType.METHOD, ElementType.TYPE})
@Retention(RetentionPolicy.RUNTIME)
@Documented
public @interface RateLimit {
/**
* 限流key，支持SpEL表达式
*/
String key() default "";
/**
* 限流算法类型
*/
Algorithm algorithm() default Algorithm.TOKEN_BUCKET;
/**
* 最大请求数
*/
int maxRequests() default 100;
/**
* 时间窗口大小（秒）
*/
long windowSize() default 60;
/**
* 令牌补充速率（个/秒）
*/
double refillRate() default 1.0;
/**
* 限流失败时的错误消息
*/
String message() default "请求过于频繁，请稍后再试";
enum Algorithm {
FIXED_WINDOW,
SLIDING_WINDOW,
TOKEN_BUCKET,
LEAKY_BUCKET
}
}
限流AOP切面
/**
* 限流AOP切面
*/
@Aspect
@Component
public class RateLimitAspect {
private final DistributedRateLimiter distributedRateLimiter;
private final SpelExpressionParser parser;
public RateLimitAspect(DistributedRateLimiter distributedRateLimiter) {
this.distributedRateLimiter = distributedRateLimiter;
this.parser = new SpelExpressionParser();
}
@Around("@annotation(rateLimit)")
public Object around(ProceedingJoinPoint joinPoint, RateLimit rateLimit) throws Throwable {
String key = generateKey(joinPoint, rateLimit);
// 检查限流
boolean allowed = false;
switch (rateLimit.algorithm()) {
case FIXED_WINDOW:
allowed = distributedRateLimiter.tryAcquireFixedWindow(
key, rateLimit.maxRequests(), rateLimit.windowSize());
break;
case TOKEN_BUCKET:
allowed = distributedRateLimiter.tryAcquireTokenBucket(
key, rateLimit.maxRequests(), rateLimit.refillRate());
break;
}
if (!allowed) {
throw new RateLimitException(rateLimit.message());
}
return joinPoint.proceed();
}
/**
* 生成限流key
*/
private String generateKey(ProceedingJoinPoint joinPoint, RateLimit rateLimit) {
if (StringUtils.hasText(rateLimit.key())) {
// 解析SpEL表达式
Expression expression = parser.parseExpression(rateLimit.key());
EvaluationContext context = new StandardEvaluationContext();
// 添加方法参数到上下文
MethodSignature signature = (MethodSignature) joinPoint.getSignature();
String[] paramNames = signature.getParameterNames();
Object[] args = joinPoint.getArgs();
for (int i = 0; i < paramNames.length; i++) {
context.setVariable(paramNames[i], args[i]);
}
return expression.getValue(context, String.class);
}
// 默认使用方法签名
MethodSignature signature = (MethodSignature) joinPoint.getSignature();
return signature.getDeclaringType().getSimpleName() + "." + signature.getName();
}
}
/**
* 限流异常
*/
public class RateLimitException extends RuntimeException {
public RateLimitException(String message) {
super(message);
}
}
🔧 实际应用示例
API限流示例
/**
* API控制器示例
*/
@RestController
@RequestMapping("/api")
public class UserController {
/**
* 用户登录 - 按IP限流
*/
@PostMapping("/login")
@RateLimit(
key = "#request.remoteAddr",
algorithm = RateLimit.Algorithm.FIXED_WINDOW,
maxRequests = 5,
windowSize = 300,
message = "登录尝试过于频繁，请5分钟后再试"
)
public ResponseEntity<?> login(@RequestBody LoginRequest request, HttpServletRequest httpRequest) {
// 登录逻辑
return ResponseEntity.ok("登录成功");
}
/**
* 发送短信验证码 - 按手机号限流
*/
@PostMapping("/sms/send")
@RateLimit(
key = "#request.phoneNumber",
algorithm = RateLimit.Algorithm.TOKEN_BUCKET,
maxRequests = 3,
refillRate = 1.0/60,
message = "短信发送过于频繁，请稍后再试"
)
public ResponseEntity<?> sendSms(@RequestBody SmsRequest request) {
// 发送短信逻辑
return ResponseEntity.ok("短信发送成功");
}
/**
* 文件上传 - 按用户ID限流
*/
@PostMapping("/upload")
@RateLimit(
key = "#userId",
algorithm = RateLimit.Algorithm.TOKEN_BUCKET,
maxRequests = 10,
refillRate = 0.1,
message = "上传过于频繁，请稍后再试"
)
public ResponseEntity<?> uploadFile(@RequestParam String userId, @RequestParam MultipartFile file) {
// 文件上传逻辑
return ResponseEntity.ok("上传成功");
}
}
限流监控和统计
/**
* 限流监控服务
*/
@Service
public class RateLimitMonitorService {
private final RedisTemplate<String, Object> redisTemplate;
private final String statsKeyPrefix = "rate_limit_stats:";
public RateLimitMonitorService(RedisTemplate<String, Object> redisTemplate) {
this.redisTemplate = redisTemplate;
}
/**
* 记录限流事件
*/
public void recordRateLimitEvent(String key, boolean allowed) {
String statsKey = statsKeyPrefix + getCurrentHour();
String field = key + (allowed ? ":allowed" : ":blocked");
redisTemplate.opsForHash().increment(statsKey, field, 1);
redisTemplate.expire(statsKey, Duration.ofDays(7));
}
/**
* 获取限流统计
*/
public Map<String, Object> getRateLimitStats(String key, int hours) {
Map<String, Object> stats = new HashMap<>();
long totalAllowed = 0;
long totalBlocked = 0;
for (int i = 0; i < hours; i++) {
String statsKey = statsKeyPrefix + (getCurrentHour() - i);
Object allowed = redisTemplate.opsForHash().get(statsKey, key + ":allowed");
Object blocked = redisTemplate.opsForHash().get(statsKey, key + ":blocked");
if (allowed != null) {
totalAllowed += Long.parseLong(allowed.toString());
}
if (blocked != null) {
totalBlocked += Long.parseLong(blocked.toString());
}
}
stats.put("totalAllowed", totalAllowed);
stats.put("totalBlocked", totalBlocked);
stats.put("totalRequests", totalAllowed + totalBlocked);
stats.put("blockRate", totalAllowed + totalBlocked > 0 ?
(double) totalBlocked / (totalAllowed + totalBlocked) : 0);
return stats;
}
private long getCurrentHour() {
return System.currentTimeMillis() / (60 * 60 * 1000);
}
}
💡 面试常见问题解答
Q1: 令牌桶和漏桶算法的区别？
标准回答：
令牌桶 vs 漏桶算法： 1. 令牌桶算法 - 允许突发流量：桶中有令牌就能处理 - 平滑限流：令牌按固定速率补充 - 适用场景：API限流、突发请求处理 2. 漏桶算法 - 强制限制输出速率：匀速处理请求 - 平滑输出：无论输入如何，输出恒定 - 适用场景：流量整形、消息队列 3. 主要区别 - 突发处理：令牌桶允许，漏桶不允许 - 实现复杂度：令牌桶较简单 - 应用场景：令牌桶更适合API限流 根据业务需求选择合适的算法。
Q2: 分布式限流如何实现？
标准回答：
分布式限流实现方案： 1. 基于Redis - 使用Lua脚本保证原子性 - 支持多种限流算法 - 性能好，易扩展 2. 基于数据库 - 使用数据库锁机制 - 一致性强但性能较差 - 适用于对一致性要求高的场景 3. 实现要点 - 原子性操作：避免并发问题 - 高可用性：Redis集群部署 - 性能优化：Lua脚本减少网络开销 - 监控告警：实时监控限流效果 Redis + Lua脚本是最常用的方案。
Q3: 如何设计一个高性能的限流系统？
标准回答：
高性能限流系统设计： 1. 算法选择 - 根据场景选择合适算法 - 令牌桶：API限流 - 滑动窗口：精确控制 2. 存储优化 - Redis集群：提高并发能力 - 本地缓存：减少网络开销 - 数据分片：避免热点问题 3. 性能优化 - Lua脚本：减少网络往返 - 批量操作：提高吞吐量 - 异步处理：避免阻塞 4. 监控运维 - 实时监控限流效果 - 动态调整限流参数 - 告警机制：及时发现问题 关键是在准确性和性能间找到平衡。
核心要点总结：
- ✅ 掌握各种限流算法的原理和实现
- ✅ 理解分布式限流的技术方案
- ✅ 能够设计基于注解的限流框架
- ✅ 了解限流系统的监控和运维要点
Java面试圣经 文章被收录于专栏
Java面试圣经，带你练透java圣经
投递美团等公司10个岗位
查看9道真题和解析
