1. Pipeline

The Redis pipeline (Pipelining) sends multiple commands at once

Save round trip time

Reduce communication costs

** Important note ** : When piping a command, the server will be forced to reply to a queue, consuming a lot of memory.

2. Publish and subscribe Pub/Sub

2.1 Application Scenarios

  • Publish and subscribe system logs
  • Sentinels find each other
  • Publish and subscribe decouple the business

2.2 Common Commands

** PUBLISH channel message**

Sends the message to the specified channel, channel

Time complexity: O(N+M)

N is the number of subscribers to channel

M is the number of clients using Subscribed patterns.

** SUBSCRIBE channel [channel …] 支那

Subscribe to information for a specified channel

Time complexity: O(N)

N is the number of channels subscribed to.

** UNSUBSCRIBE [channel [channel …]]**

Instructs the client to unsubscribe from a given channel or all channels if no channel is specified

Time complexity: O(N)

N is the number of channels subscribed to.

** PSUBSCRIBE pattern [pattern …] 支那

Subscribe to a given pattern pattern is a match to a channel

Time complexity: O(N)

N is the number of subscribed modes.

** PUNSUBSCRIBE [pattern [pattern …]] **

Instructs the client to unsubscribe from the specified schema or exit all schemas if no schema is provided

Time complexity: O(N+M)

N is the number of schemas subscribed by the client

M is the number of modes subscribed by all clients in the system.

** PUBSUB subcommand [argument [argument …]] **

PUBSUB is an introspective command that looks at the status of a subscription and publishing system. It consists of several subcommands in different formats

  • PUBSUB CHANNELS [pattern]

Lists the current active Channels. Active means the channel has one or more subscribers

Responds only to SUBSCRIBE UNSUBSCRIBE commands

  • PUBSUB NUMSUB [channel-1 … channel-N]

Lists the number of subscribers for the specified channel

Responds only to SUBSCRIBE UNSUBSCRIBE commandsCopy the code
  • PUBSUB NUMPAT

Returns the number of subscription patterns

Responds only to the PSUBSCRIBE PUNSUBSCRIBE command

3. The transaction

A set of commands is processed in the same transaction.

Redis is single-process, if there are more than one client sent instructions are queued to execute.

Redis transactions do not support rollback and, unlike database transactions, remain simple and fast

When a transaction is executed, the client whose EXEC command arrives first will be executed first, as shown in the following figure:

3.1 Common Commands

** MULTI **

Open the transaction

** EXEC **

Perform transactions

With WATCH, EXEC is executed only if the monitored key is not modified and the setup mechanism is allowed to be checked

** DISCARD **

Abort transaction execution

DISCARD releases all WATCH keys when a WATCH is present

** WATCH key [key …] 支那

Flags that all specified keys are monitored for conditional execution in a transaction (optimistic locking)

** UNWATCH **

Unmonitor all keys with the WATCH command

If you perform EXEC or DISCARD, you do not need to manually execute UNWATCH

Four Modules – RedisBloom

RedisBloom extension program, provides 4 kinds of data structures, Bloom Filter, a cuckoo filter, a count-min Sketch, a top-K.

Bloom and Cuckoo filter are used to determine whether an element is an element of a set, how it works

A count-min Sketch is used as a frequency table for events in the data stream, and the hash function maps events to frequencies

A top-k is used to maintain the k most frequently occurring sequences

4.1 Bloom Filter

Use bloom filters to reduce disk IO or network requests

** Note ** when using Bloom Filter

  1. It’s penetrating. It doesn’t exist
  2. Client, add key and value tags in Redis
  3. The database adds elements
  4. Finish adding the element to bloom

Bloom shortcomings

  1. The query performance is relatively weak because it uses hash functions to jump from point to point on the bitmap, making it difficult to exploit the CPU cache
  2. Deletion not supported

4.2 cuckoo Filter

Cuckoo filters are not implemented using bitmaps, but one-dimensional arrays. What it stores is a fingerprint of the data











Refer to websites and pages

redis.cn/

redis.io/

Blog.csdn.net/cs958903980…