redis database

redis.md

Data Types

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

• binary-safe strings
• lists: strings (insertion order) (linked list instead of array)
• sets: strings
• sorted sets: (strings, score)
• hashes: {string:string}
• bit array: bitmaps
• HyperLogLogs
• streams: append-only map-like

Transactions

https://redis.io/topics/transactions

(Redis doesn't support roll backs. Even some cmds failed, it will execute the rest.)

• MULTI: enter a transaction
• EXEC: execute cmd
• DISCARD: flush the transaction queue (no cmds are executed)
• WATCH: check-and-set (CAS)

High Availability

https://redislabs.com/redis-enterprise/technology/highly-available-redis/

• Replication
  • store dataset
  • failover
  • serve asa a tiebreaker
• Auto failover
  • node watchdog
  • cluster watchdog
• Multi availability zone

Distributed Locks

https://redis.io/topics/distlock

Single instance: SET resource_name my_random_value NX PX 30000 (not exist, expire)

Redlock:

• get the current time in milliseconds
• try to aquire the lock in all the N Redis master sequentially with a small timeout
• the time elapsed is less than lock validity time and aquired lock in the majority instances
• validity time = initial validity time - time elapsed
• try to unlock all the instance if failed, and retry after a random delay

Redis Strings

https://redis.io/topics/internals-sds

struct sdshdr {
    long len;
    long free;
    char buf[];
};

Key Expire

https://redis.io/commands/expire

• passive: expire when access the key
• active: (10 times per second)
  • test 20 random keys
  • delete all the expired keys
  • if more than 25%, start again

Persistence

https://redis.io/topics/persistence

• RDB: point-in-time snapshots of dataset at specified intervals
  • backup
  • compact
  • maximize performance (child process will do the disk I/O)
  • allow faster restarts
  • loss data
  • fort() can be time consuming
• AOF: log every write operation that will be played again at server startup
  • durable (no fsync, fsync every second, fsync at every query)
  • append only
  • automatically rewrite a minimal set of operations needed to create the current data set
  • log format is easy to understand and parse
  • bigger than RDB
  • can be slower

LRU cache

https://redis.io/topics/lru-cache

approximated LRU algorithm