redis

k-v

date type for v

https://redis.io/topics/data-types-intro

  1. string
  2. list: 队列,阻塞访问BRPOP, BLPOP
  3. set: set
  4. sorted set: for i in this: i ~= (score, v), i is sorted by score in this
  5. hash: dict with string limited?
  6. Bit arrays: 2进制位图，第i位(0,1)表示: i is False or True; 例如统计一年多少天登录过需要366bit
  7. HyperLogLogs: # example http://content.research.neustar.biz/blog/hll.html # explain http://antirez.com/news/75
    1. list=64k # 分2^^14组， 50位比特数中，记录到尾部连续(连续为正面m-1=1) # m-1个全0的为(1/2^^(m-1))，就概率有2^^m -1个不同已出现
    2. h = hash(string) 64bit 分割14bit(其10进制数对应list.index),50bit(最右边非0位**2对应list.value)
    3. 倒数平均数(set)
  8. Streams: append only list

安装

单机

pass

cluster

https://redis.io/topics/cluster-tutorial cluster-enabled yes 同kafaka, 采用选举模式, 除了学要使用命令开启过程 redis-cli –cluster create 127.0.0.1:7000 127.0.0.1:7001 \ 127.0.0.1:7002 127.0.0.1:7003 127.0.0.1:7004 127.0.0.1:7005 \ –cluster-replicas 1

同步使用异步、同步写

config

maxmemory: peak memory usage

使用

  1. base https://redis.io/commands
  2. mq https://redis.io/topics/pubsub
  3. Transactions https://redis.io/topics/transactions
  4. loadbalance/hashring https://redis.io/topics/partitioning
  5. 分布式锁逻辑
    1. 单点逻辑: clinet.set(lock.name, clinet.uuid, only_not_exist) ? ( do something done; client.get(lock.name) == client.uuid ?return del(lock) : return 0 ) : return -1
    2. 多点逻辑: 多数派逻辑(长用于选举) 依次按1, 获取各个节点锁，过一半即可认为获取到; else del all if lock.name.v == client.uuid
  6. persistence https://redis.io/topics/persistence RDB: 快照 AOF: 类似mysql, 可选多久将记录同步到磁盘