一、架构演进与核心价值
1.1 性能对比数据
1.2 协同设计优势
二、实战案例:电商商品详情页优化
2.1 痛点分析
原始架构:单层 Redis 缓存
问题现象:
# 压测数据
Requests/sec: 58000
99% latency: 120ms
Redis CPU Usage: 85%
2.2 架构改造方案
核心依赖(Spring Boot)
<!-- Caffeine依赖 -->
<dependency><groupId>com.github.ben-manes.caffeine</groupId><artifactId>caffeine</artifactId>
</dependency><!-- Redis依赖 -->
<dependency><groupId>org.springframework.boot</groupId><artifactId>spring-boot-starter-data-redis</artifactId>
</dependency>
缓存配置
@Configuration
public class CacheConfig {// 本地缓存配置(Caffeine)@Beanpublic CacheManager caffeineCache() {return Caffeine.newBuilder().expireAfterWrite(5, TimeUnit.SECONDS) // 短周期防穿透.maximumSize(1000).build();}// 分布式缓存配置(Redis)@Beanpublic RedisCacheManager redisCache(RedisConnectionFactory factory) {return RedisCacheManager.builder(factory).expireAfter(30, TimeUnit.SECONDS).build();}
}
服务层实现
@Service
public class ProductService {@Cacheable(value = "product", key = "#id", cacheManager = "caffeineCache")public Product getProduct(Long id) {return redisTemplate.opsForValue().get("product:" + id);}@CachePut(value = "product", key = "#result.id", cacheManager = "redisCache")public Product updateProduct(Product product) {// 更新数据库和Redisreturn product;}
}
三、一致性保障策略
3.1 缓存更新模式
// 双写策略
public void updateProductStock(Long productId, int delta) {// 1. 更新数据库productMapper.updateStock(productId, delta);// 2. 更新RedisredisTemplate.opsForValue().set("product:" + productId, updatedProduct);// 3. 逐出Caffeine缓存caffeineCache.evict(productId);
}
3.2 穿透/雪崩防护
// 空值防护策略
public Product getProduct(Long id) {Product product = caffeineCache.getIfPresent(id);if (product == null) {product = redisTemplate.opsForValue().get("product:" + id);if (product == null) {product = DB.get(id);if (product == null) {product = NULL_PRODUCT; // 特殊空对象}redisTemplate.opsForValue().set("product:" + id, product, 60, SECONDS);}caffeineCache.put(id, product);}return product == NULL_PRODUCT ? null : product;
}
四、性能压测与调优
4.1 压测结果
4.2 JVM 参数优化
# 优化GC配置
-XX:+UseG1GC
-XX:MaxGCPauseMillis=100
-XX:InitiatingHeapOccupancyPercent=35
五、落地 Checklist
5.1 监控体系
# Prometheus监控指标
-pattern:"cache_hits_total"
name:"caffeine_cache_hits"
labels:tier:"local"
-pattern:"redis_cache_hits"
labels:tier:"distributed"
5.2 降级策略
// 降级开关
@HystrixCommand(fallbackMethod = "getProductFallback")
public Product getProduct(Long id) {// 正常逻辑
}public Product getProductFallback(Long id) {return localCache.getIfPresent(id); // 降级到本地缓存
}
5.3 冷启动策略
// 预热脚本
@PostConstruct
public void warmUpCache() {List<Long> hotIds = productMapper.selectHotIds(1000);hotIds.parallelStream().forEach(id ->caffeineCache.put(id, productMapper.selectById(id)));
}
六、总结与思考
架构设计三原则:
- 速度优先:Caffeine 处理 80%热数据
- 容量分层:Redis 存储全量数据
- 最终一致:通过 MQ 实现异步同步
未来演进: