What is the b+tree index of MySQL? How does the cluster index grow?

​ You must know MySQL Conduct CRUD It's in memory , That is to say Buffer Pool in . And then you know when there's no memory MySQL When data is needed ,MySQL From Disk Pass through IO Operation to read data into memory . The reading unit is ： Data pages

The general data page length is as follows

​

you 're right , Real data is stored in the data page , And the data pages are organized in the form of two-way linked tables in memory ！ Here's the picture

​

And in the B+Tree Setting of , It requires that the primary key index be incremented , That is to say, if the primary key index is incremented , All data in the data page on the right is required to be larger than that in the data page on the left . But it's clear that the picture doesn't fit , So you need to split the page to make it look like this .

​

good , Now think about it , You must have heard of it before ：MySQL Of B+Tree Cluster index , Only leaf nodes store real data , Instead of storing index data in leaf nodes , And leaf nodes are connected by two-way linked list

you 're right , All that B+Tree The leaf node of is the data page in the figure above , And they are really linked through a two-way linked list ！

Let's move on , If you only look at the two-way list connected by data pages in the figure above , Now if we search id=7 The data line , What will happen then ？

Obviously we're going to scan from the beginning ！

Then you may ask ： I didn't mean B+Tree The primary key is required to be incremental ？ And there is a page splitting mechanism to ensure that all data in the right data page is larger than the index value of the left data page . It's not good to do a binary search ？

answer ： Yes , It's really possible to do a binary search in a single data page , But the organization relationship between data pages is linked list , So traversal from the beginning is inevitable .

that MySQL What to do ？

Here's the picture ：MySQL An index directory is abstracted for many data pages

With this index, we can easily search through many data pages ！ Have the ability of binary search directly ！

And this directory actually exists in the data page , Unlike leaf nodes , It only stores index information , The leaf node stores real data ？

And the birth of index pages means B+Tree The rudiment of the world has been born ！

As users continue to select,buffer pool There are more and more data pages in , Then the data in the index page will also rise . When the existing index volume exceeds 16KB（ The capacity of a data page ） We have to create a new index page to store new index information . And then this one B+Tree It's going to get fatter and fatter .

Then you know B+Tree yes B Varieties of trees , and B Trees can actually be 2-3 Trees 、2-3-4 Count .... wait M A general term for rank trees , When the number of elements that can be stored in each node reaches the upper limit , The tree will grow high .

Like the picture below ：

​