Strong! I got 10w pageviews in 3 days after going online, JD T8 Pure Hand Code Redis Caching Manual, I fan

Hello, today I will share the Redis cache with all the children's shoes, please take out a small book and write it down!

Introduction

What is Redis?

Redis is a storage system, like MongoDB, it is a key-value pair storage, that is, a key-value structure. In other words, Redis is actually a kind of database, which can be used across platforms. This series integrates SpringBoot+Redis.

Redis itself is written in C language and is an open source system that complies with ANSI C standards. Redis is the abbreviation of Remote Dictionary Server, and is now commonly used in distributed databases and high-availability cache systems. In fact, after learning Redis to integrate SpringBoot, integrating SpringCloud is a matter of course.

Why use Redis?

Some friends may ask, since Redis and MongoDB are so similar, can I just use the latter? Why should I learn from the former? In fact, MongoDB and Redis are very different, they just have similarities in storage. The differences between the two are listed below:

• Data storage location

The data stored in MongoDB is stored on disk, and a small part of the data is stored in memory. Why should the database data be stored in memory? It is because the cache system of the database will judge the hot data, and the frequently accessed data must be stored in the memory for efficient processing.

Redis data is all stored in memory and written to disk regularly, thus creating the high performance of Redis. When the memory is not enough, Redis will use the LRU (Least Recently Used) algorithm to replace the data, yes, it is the west replacement algorithm of the operating system.

• Data storage method

MongoDB uses the mmap function to add, delete, and modify the memory mapped by the file data. After the modification, the operating system flushes the memory data to the disk. The problem is that the two are not one transaction, so if there is a downtime between the two transactions, the data will naturally be lost.

Redis has two storage modes, RDB and AOF modes, which we will talk about later.

• Data processing speed

MongoDB is slower than Redis, which is also the biggest advantage of Redis as a memory storage system. Of course, we do not consider the situation of insufficient memory for the time being.

In addition, currently distributed is the mainstream technology stack of major manufacturers, and Redis is the distributed caching method that major manufacturers will adopt. And in dealing with high concurrency scenarios, such as Ali's Double Eleven, the appearance of new Tencent game skins or Chinese New Year red envelopes, the use of Redis and message queues is the mainstream method.

What is the level of entry? What is the level of proficiency?

Getting started is to have no knowledge of Redis, but you need to have basic knowledge, such as basic data structure, basic type, basic memory knowledge, basic distributed knowledge,

Basic knowledge of Java/Spring and basic operation of Linux. You can rest assured that my level is average, so the foundation I said must be the foundation.

Being proficient does not mean that I can finally write a Redis by myself, or that I can become a stable developer and maintainer of Redis. I can't do it myself, so the proficiency here is to be able to understand the various mechanisms and algorithms of Redis, and to understand the common uses provided by Redis. Method, understand the usage scenarios of Redis, can pit holes by yourself, and be able to integrate the SpringBoot framework by yourself. In fact, this integration of the SpringBoot framework is the simplest, and the official website is very detailed. It is best to carry out secondary development on the basis of Redis. In fact, after understanding the principle and knowing how to write C language, secondary development is not difficult.

Since it is a series of articles, I cannot write all of them in one article. Too much content is not good for my typesetting and learner experience, so I will try to compress each article within an acceptable length range. Of course , I try to give a link to all articles in the series in every article for everyone and for my convenience in the future. If the article can help you, I hope to give a compliment and encouragement. Although I don't live by this, I'm very happy to be recognized.

Caching and usage scenarios

What is caching?

Caching is actually a way for a program to use memory to optimize frequently read data. The cache is part of the memory. We wrote earlier that database operations are often stored in disks, but disk IO is extremely slow. What should I do? The programmers thought of a way to make good use of memory. Isn't it because the disk IO is slow, then use the memory, the memory IO is fast, but there are too many things to put in the memory, what should I do? Then put the frequently used data in the memory.

It is 2021, and friends who are familiar with the NBA know that Westbrook’s data exploded again this year, so fans may often look at Westbrook’s data. In addition, there are many fans of Westbrook, and there may be a problem of high concurrency. But it is too slow to get data from the disk every time, so put Westbrook's data in the memory separately, and other people's data continue to lie in the disk, which greatly speeds up the system's response speed. This is the cache.

However, there is more than one star in the NBA. Curry, James, Jokic, etc. are all excellent players. The number of fans is no less than Westbrook, so put all their data in the cache, and it will be fast. . But we found that there are too many NBA stars to store inside. What should we do? This leads to our common cache elimination algorithm. Friends who know the operating system may know what LRU, LFU, FIFO, FILO, etc., I will put this in the Redis algorithm mechanism behind.

Local cache and distributed cache

Let’s take a look at the example. It’s still an NBA player. My memory is too small. Each computer can only store the data of one NBA player. But what should I do if I have three players’ data that need to be stored in the computer?

Local cache

Local cache is to store data in local memory. Because it does not need to connect to other hosts through the network, it is naturally the fastest, and of course there are disadvantages. For example, Mybatis primary and secondary cache, Caffeine, Guava are all typical examples of local cache. Going back to the previous example, 3 players are placed on 3 different hosts. If you use a local cache architecture, this is how:

advantage:

Fast speed, no need to transmit over the network

Disadvantages:

Limited available cache capacity per host

Multiple nodes cannot share data

Distributed cache

Distributed cache is to use the network to put the cache on a certain host, so that the capacity limit of the cache is the memory size of the cache machine, and the IO bottleneck is the network speed. Redis is a typical distributed cache. Of course, we can also make Redis a complete local cache without network calls.

Of course, the above picture is not accurate. Since the cache is distributed, of course, it cannot be deployed on only one cache machine. The cache machines are often deployed in a cluster.

Hot key issues

What is a hot key? Quite simply, we see that Weibo often goes down. In addition to the large number of users and the amount of information, it is often a celebrity who announces "We have a child" and other things. Countless users visit this resource, and this resource corresponds to it. The key is called a hot key. Suddenly high concurrency makes a certain hotspot frequently accessed, which leads to exhaustion of server cache resources. How to deal with it?

There are many solutions, but today we talked about the caching problem, just talk about the caching solution. Take the example of a star having a child, from the analysis of the cause to the solution, in fact, the reason is there, and the solution is just a matter of hand.

Reason analysis: A certain celebrity A and a certain celebrity B have children, and A and B are on the distributed cache server, but when this kind of explosive news happens, a large number of users access the distributed cache server, even if the cache resources are not exhausted, the network IO cannot be carried either. Since our assumption is that the key is on one machine, we don't consider splitting the key for the time being, that is, the cache can withstand it temporarily, but the network IO has collapsed, and the distributed cache server cannot withstand such a large amount of data. What should I do?

Solution: Since the network problem can't be solved, it won't be solved, just don't use it. After all, the method that layman knows to expand network bandwidth can't enter our eyes. Why not? That is local cache + distributed cache. If the network can't stand it, it puts high-frequency resources on the local server, and the outside world accesses related resources directly from the local server, which has nothing to do with the distributed cache server. Only when new high-frequency resources are accessed and the old high-frequency resources continue to be hot, replace it.

Alright, that’s all for today’s article, I hope to help you who are confused in front of the screen