架构师视角 | 分布式缓存如何选择 ?

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如今，缓存系统的应用非常广泛，能够用来提高并发数、数据吞吐量，提高快速响应能力。那么当数据量达到一定程度，单机环境可能就显得有些力不从心了，就需要一个分布式缓存系统。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"1. 缓存系统的选择","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/d2/d2a6ab302e0ab5c7302ebfc95f76520f.png","alt":"","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"1.1 缓存分类","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如上图所示，首先缓存大致可以分为四大类。","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"CDN 缓存：CDN 即内容分发网络，CDN 边缘节点将数据缓存起来。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"反向代理缓存：如 Nginx 的缓存。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"本地缓存：代表的有 EhCache 和 Guava Cache。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"分布式缓存：各缓存系统。","attrs":{}}]}],"attrs":{}}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"1.2 分布式缓存","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"本文主要探讨各分布式缓存系统，如图 1-1 所示，列出了五种：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"其中 EvCache 和 Aerospike 使用场景不是那么通用和广泛。","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"EvCache：是 Netflix 的基于 Memcached & Spymemcached 的缓存方案。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Aerospike：是可基于 SSD 的 KV NoSQL 数据库。","attrs":{}}]}],"attrs":{}}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"除此之外，还有三种常见缓存系统。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/a8/a8beb9ad9ddbe9458aaeed256b0f7429.png","alt":"","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Tair：阿里开源，跨机房、性能随结点添加线性上升、适用大数据量。Tair 还有三种引擎。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"LDB: 基于 google levelDB，支持 KV和类 HashMap 结构，性能稍低，持久化可靠性最高。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"MDB: 基于 Memcache，支持 KV 和类 HashMap，性能最优，不支持持久化存储。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"RDB: 基于 Redis。","attrs":{}}]}],"attrs":{}}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Memcache: 不支持数据同步、分布式支持较差。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Redis: 社区活跃、使用最多。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"综上所述，在一般情况下，考虑到适用性和稳定性，Redis 是搭建缓存系统的最优选择。以下将基于 Redis 介绍。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"2. Redis 集群缓存方案","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如顶部图 1-1 所示，列出了 Redis 的集群高可用的方案，基本可以分为三种。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2.1 主从机制","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"常见的集群架构，搭建简单，主要实现读写分离和备份，可以由 Master 负责读写，Slave 负责备份。但存在故障恢复复杂、水平拓展难、写能力受限等问题。结构图如下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/b1/b136679b1303df46347c5a57ce0960e0.png","alt":"","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2.2 哨兵机制","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Redis Sentinel 是社区版本推出的原生高可用解决方案。由一或多个哨兵实例监视任意个主从服务器，且在 Master 宕机时，自动将宕机服务器属下的 Slave 服务器升级为 主服务器，从而保证系统的可用性。较主从实现的监控、选主。但问题主要是要保证 Master 的 HA 切换。结构图如下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/fa/fa33d8cafc52d2c4ecee64ad6c7058a6.png","alt":"","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2.3 \"分布式\"","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"到这里以上两种机制其实只能算作“集群”，并非严格意义上的“分布式”。接着来看看分布式方案。","attrs":{}}]},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"集群强调高可用，分布式在集群的基础上又强调协作。","attrs":{}}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"3. Redis分布式缓存方案","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"任何分布式存储系统，首先面临的就是 sharding（分片）问题，如顶部图 1-1 所示该问题有为三种解决方法。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"3.1 客户端分片","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"顾名思义，将数据分片的路由功能交给客户端，但这是一种静态分片，维护性差。基本是不予考虑的。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"3.2 代理分片","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"通过代理分发到具体的 redis 实例。有两个常用解决方案。","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Twemproxy：Twitter 开源，轻量级，不再维护，无法平滑地扩容/缩容，运维也不是很友好，性能一般。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Codis：豌豆荚开源，支持水平拓展，运维平台完善，性能较 Twemproxy 快。Codis 在国内使用的较多，同时代理分片的思路也有很多公司在此基础开发了自己的二次方案。不过 Codis 也不再维护。","attrs":{}}]}],"attrs":{}}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"其实，这两种代理分片的方案，都是在 Redis 官方未推出良好的分布式方案时的产生的，在官方更新提供更优策略后都不再维护。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"3.3 服务器端分片","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这就要谈到 Redis 官方方案 Redis-cluster 。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在 Redis 3.0 之前是没有较好的分布式方案的，这也是第三方方案出现的原因。3.0 开始，官方推出了去中心化的分布式方案。集群中包含 16384 个散列槽，每个节点负责其中一部分。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"先看下拓扑图：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/49/493ce1fb5e40cb7b200f808db253c42d.png","alt":"","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"每个节点打开两个 TCP 连接，一个负责给客户端提供服务，一个负责节点间通信。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此刻要说说 CAP 了 ：Consistency（一致性）、Availability（可用性）、Partition tolerance（分区容错性） 。对分布式系统而言，CAP 必须牺牲一者。Redis Cluster 的设计目标主要是高性能、高可用和高扩展，只好抛弃一部分数据一致性。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"数据一致性：由于Redis Cluster 使用异步复制， 在某些情况下如 Master 宕机但未同步至 Slave，可能会导致丢失写入。在绝对需要支持同步写入时，可通过 WAIT 命令实现，可使得丢失写入的可能性大大降低。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"可用性：当集群中一部分节点故障后，集群整体能响应客户端读写请求。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/66/66b5ff020152927fae87c76dfbff218c.png","alt":"","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"节点间定时互 ping ，当超过一半 Master 判定某节点失败，则标记为 FAIL，且会向集群广播节点下线的消息。如下线节点是带有槽的主节点，则要从它的从节点选出一个替换。","attrs":{}}]}],"attrs":{}}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"高性能和拓展：操作某个 key 时，不会先找到节点再处理，而是直接直接重定向到该节点，同时相较代理分片也少了 proxy 的连接损耗。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"但是在进行 multiple key 操作时需要 keys 位于同一个 slot 上，需要使用 hash tags，使用 {} 强制将某些 key 映射到每个 slot，以便进行 multiple 。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在拓展方面，Redis Cluster 最大支持线性拓展 1000 个节点，将新节点加入集群后可以通过命令指定和平均的从已有节点分配 slot。","attrs":{}}]}],"attrs":{}}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"4. 缓存常见问题","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"以上介绍了简单介绍了常见缓存系统，并具体列出了基于 Redis 的集群方案。下面谈一谈缓存系统常见的问题。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如下图所示，列出七个常见问题。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/cc/ccdc0c31c295c9243fc144ed5b02d6a4.png","alt":"","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"4.1. 缓存穿透","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"指访问不存在的数据，从而绕过缓存，直接请求到了数据源，当请求过多，就会对 DB 造成压力。","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"空 key：指对于不存在的数据也将 key 存空值入缓存系统，这样下次访问也会得到返回。但只适用于空数据 key 有限、key 重复请求概率高，如果量大且不重复，就会造成很多无用 key 的创建。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"布隆过滤器：布隆过滤器是一个很长的二进制向量和一系列随机映射函数。可用于检索一个元素是否在一个集合中加一层对空值的过滤器，空间和时间效率都很高。但由于 hash 产生的碰撞可能存在误判，以及因不存储 key 导致的无法删除。适用于空数据 key 各不同、重复请求概率低。","attrs":{}}]}],"attrs":{}}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"4.2. 缓存击穿","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"缓存击穿实际是缓存雪崩的一个特例。指当某些热点 key 过期时，就会有大量的请求击穿到 DB。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"互斥锁：在缓存失效的时候，不立即 load db，可以先用如 SETNX 等命令去 set 一个 mutex key，当操作返回成功时，说明拿到锁，此刻该线程进行 load db 的操作并更新缓存；否则未拿到锁就（可休眠一段）重试 get 缓存的方法。但要注意死锁风险。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"不过期","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这里的不过期有两个概念，一个指未设过期时间，那是真的不过期，那没事了。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"另一个是指通过业务逻辑，将 key 的过期时间进行存储，请求是判断是否小于值，是则后台异步更新。","attrs":{}}]}],"attrs":{}}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"4.3. 缓存雪崩","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"同一时刻大量缓存失效（故障）, 请求到了 DB。","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"随机时间：在设置过期时间时，可以在基础时间上 + 一个随机的时间，等于实现了分批过期。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"后台更新：将更新失效的工作交给后台定时线程。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"限流 + 本地缓存：如 ehcache 本地缓存 + Hystrix 限流。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"双缓存：类似于设置主从缓存，从 key 不过期。","attrs":{}}]}],"attrs":{}}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"4.4. 缓存更新与一致性","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果保证数据一致性。列出四种更新策略：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Cache Aside ：最常用的。失效时回源取数据，更新；命中时，返回缓存数据；更新时先数据源更新，再更新缓存。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Write Back ：更新数据时，只更新缓存，不更新数据源。缓存异步批量更新数据库。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Read/Write Through","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Write Through ：当有数据更新时，如未命中缓存，直接更新数据库，并返回。如命中缓存，则更新缓存，再由 Cache 自己更新数据库。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Read Through ：更新数据源由缓存系统操作，读取数据时如缓存失效，则取回源数据更新缓存。","attrs":{}}]}],"attrs":{}}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"4.5. 热点数据","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"对于热点数据的处理方法。","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"拆分复杂结构：如二级数据结构，进行拆分，这样热点 key 就被拆为若干个的 key 分布到不同节点。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"迁移热点：对于 Redis Cluster 而言可以将热点 key 所在的 slot 单独迁移到一个节点，降低其他节点压力。","attrs":{}}]}],"attrs":{}},{"type":"listitem","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"多副本：复制多份缓存副本，将请求分散到多个节点上，减轻单台缓存服务器压力，适合多读少写。","attrs":{}}]}],"attrs":{}}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"4.6. 缓存预热","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"指可以将某些的缓存数据提前加载到缓存系统，提前避免在如热点数据大量请求到库。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"4.7. 缓存降级","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"指当访问量剧增、服务出现问题或非核心服务影响到核心流程的性能时，仍需保证主服务可用。可根据一些关键数据自动降级，也可配置开关人工降级。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"5. Redis Cluster 使用","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"对于 Redis Cluster 环境的搭建和基础使用非常简单。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"无论基于何种方式，只要搭建好 n 台 redis 服务并保证各服务间可以互相通讯后，任意进入一个 redis 服务键入：","attrs":{}}]},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"redis-cli --cluster create IP1:port1 IP2:port2 IP3:port3 IP4:port4 IP5:port5 IP6:port6 ... --cluster-replicas 1","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"即可。之后可以使用 cluster node 和 cluster info 命令查看集群、节点信息。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而对于广大 JAVA 开发，Spring Data Redis 从 1.7 起即支持 Redis Cluster，只需配置 Master 节点地址（和密码）。","attrs":{}}]},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"spring.redis.cluster.nodes=ip1:port1,ip2:port2,ip3:port3","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"加入依赖","attrs":{}}]},{"type":"codeblock","attrs":{"lang":null},"content":[{"type":"text","text":"compile(\"org.springframework.boot:spring-boot-starter-data-redis\")","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"即可通过 RedisTemplate 使用。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"6. 总结","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"本文从缓存系统的选择出发，基于 Redis 介绍了几种集群方案并重点说明了 Redis Cluster 方案。之后列出缓存系统常见问题及常见解决方案，最后对使用做了简单说明。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"当然，如何去落地，如何解决这些问题还需要根据实际场景具体分析和处理。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https://www.bilibili.com/video/BV17z4y1k7zF/","title":null},"content":[{"type":"text","text":"MySQL从入门到进阶教程，主讲老师：马士兵、连鹏举","attrs":{}}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https://blog.csdn.net/Java0258/article/details/109403289?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522160665848319195265186288%2522%252C%2522scm%2522%253A%252220140713.130102334.pc%255Fall.%2522%257D&request_id=160665848319195265186288&biz_id=0&utm_medium=distribute.pc_search_result.none-task-blog-2~all~first_rank_v2~rank_v28-1-109403289.pc_first_rank_v2_rank_v28&utm_term=%E5%AD%97%E8%8A%82%E8%B7%B3%E5%8A%A8%E6%80%BB%E7%BB%93%E7%9A%84%E8%AE%BE%E8%AE%A1%E6%A8%A1%E5%BC%8F%20PDF%20%E7%81%AB%E4%BA%86%EF%BC%8C%E5%AE%8C%E6%95%B4%E7%89%88%E5%BC%80%E6%94%BE%E5%88%86%E4%BA%AB&spm=1018.2118.3001.4449","title":null},"content":[{"type":"text","text":"字节跳动总结的设计模式 PDF 火了，完整版开放分享","attrs":{}}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https://blog.csdn.net/javachengzi/article/details/109729058?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522160665850419721940242314%2522%252C%2522scm%2522%253A%252220140713.130102334.pc%255Fall.%2522%257D&request_id=160665850419721940242314&biz_id=0&utm_medium=distribute.pc_search_result.none-task-blog-2~all~first_rank_v2~rank_v28-1-109729058.pc_first_rank_v2_rank_v28&utm_term=%E5%88%B7Github%E6%97%B6%E5%8F%91%E7%8E%B0%E4%BA%86%E4%B8%80%E6%9C%AC%E9%98%BF%E9%87%8C%E5%A4%A7%E7%A5%9E%E7%9A%84%E7%AE%97%E6%B3%95%E7%AC%94%E8%AE%B0%20%E6%A0%87%E6%98%9F70.5K&spm=1018.2118.3001.4449","title":null},"content":[{"type":"text","text":"刷Github时发现了一本阿里大神的算法笔记！标星70.5K","attrs":{}}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https://blog.csdn.net/javachengzi/article/details/108219072","title":null},"content":[{"type":"text","text":"如果能听懂这个网约车实战，哪怕接私活你都可以月入40K","attrs":{}}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https://blog.csdn.net/javachengzi/article/details/108219072","title":null},"content":[{"type":"text","text":"为什么阿里巴巴的程序员成长速度这么快，看完他们的内部资料我懂了","attrs":{}}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https://www.bilibili.com/video/BV1f541157DX?spm_id_from=333.788.b_765f64657363.2","title":null},"content":[{"type":"text","text":"程序员达到50W年薪所需要具备的知识体系。","attrs":{}}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https://www.bilibili.com/video/BV1Ay4y167CV/","title":null},"content":[{"type":"text","text":"关于【暴力递归算法】你所不知道的思路","attrs":{}}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"看完三件事❤️","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果你觉得这篇内容对你还蛮有帮助，我想邀请你帮我三个小忙：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"点赞，转发，有你们的 『点赞和评论』，才是我创造的动力。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"关注公众号 『 Java斗帝 』，不定期分享原创知识。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"同时可以期待后续文章ing🚀","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}}]}