mysql advanced three: sql performance optimization + index optimization + slow query log

Introduction

Single table index failure case

0. Thinking question: How to improve insertion efficiency if 1 million data is inserted into MYSQL

(1) Turn off automatic submission and only submit once manually

(2) Delete other indexes except the primary key index

(3) Spell the long sql that mysql can execute, and insert data in batches

(4) Use java multithreading

(5) Use the framework, set attributes, and implement batch insertion

1. Calculations and functions lead to index failure

CREATE INDEX idx_name ON emp (NAME);

EXPLAIN SELECT * FROM emp WHERE emp.name LIKE 'abc%';

EXPLAIN SELECT * FROM emp WHERE LEFT(emp.name,3) = 'abc'; ----Index failure

2 LIKE index starting with % is invalid

EXPLAIN SELECT * FROM emp WHERE NAME LIKE '%ab%'; ----Index failure

3. Not equal to (!= or <>) index failure

EXPLAIN SELECT * FROM emp WHERE emp.name = 'abc' ;

EXPLAIN SELECT * FROM emp WHERE emp.name <> 'abc' ; ----Index failure

4、IS NOT NULL 和 IS NULL

EXPLAIN SELECT * FROM emp WHERE emp.name IS NULL;

EXPLAIN SELECT * FROM emp WHERE emp.name IS NOT NULL; ----Index failure

5. Type conversion causes index failure

EXPLAIN SELECT * FROM emp WHERE NAME='123';

EXPLAIN SELECT * FROM emp WHERE NAME= 123; ----Index failure

6. Full value matches my favorite

EXPLAIN SELECT * FROM emp WHERE emp.age = 30 AND deptid = 4 AND emp.name = 'abcd';

CREATE INDEX idx_age ON emp(age);

CREATE INDEX idx_age_deptid ON emp(age,deptid);

CREATE INDEX idx_age_deptid_name ON emp(age,deptid,`name`);

7. Best Left Prefix Rule

EXPLAIN SELECT * FROM emp WHERE emp.age=30 AND emp.name = 'abcd' ;

CREATE INDEX idx_age_name ON emp (age,NAME);

EXPLAIN SELECT * FROM emp WHERE emp.deptid=1 AND emp.name = 'abcd';

EXPLAIN SELECT * FROM emp WHERE emp.age = 30 AND emp.deptid=1 AND emp.name = 'abcd';

CREATE INDEX idx_age_deptid_name ON emp(age,deptid,`name`);

EXPLAIN SELECT * FROM emp WHERE emp.deptid=1 AND emp.name = 'abcd' AND emp.age = 30;

8. The column on the right side of the range condition in the index is invalid

CREATE INDEX idx_age_deptid_name ON emp(age,deptid,`name`);

EXPLAIN SELECT * FROM emp WHERE emp.age=30 AND emp.name = 'abc' AND emp.deptId>1000 ;

CREATE INDEX idx_age_name_deptid ON emp(age,`name`,deptid);

Association query optimization

1. Data preparation

-- 分类

CREATE TABLE IF NOT EXISTS `class` (

`id` INT(10) UNSIGNED NOT NULL AUTO_INCREMENT,

`card` INT(10) UNSIGNED NOT NULL,

PRIMARY KEY (`id`)

);

-- 图书

CREATE TABLE IF NOT EXISTS `book` (

`bookid` INT(10) UNSIGNED NOT NULL AUTO_INCREMENT,

`card` INT(10) UNSIGNED NOT NULL,

PRIMARY KEY (`bookid`)

);





-- 插入16条记录

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO class(card) VALUES(FLOOR(1 + (RAND() * 20)));



-- 插入20条记录

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

INSERT INTO book(card) VALUES(FLOOR(1 + (RAND() * 20)));

2、左外连接实例

（1）明确角色

（2）优化

EXPLAIN SELECT * FROM class LEFT JOIN book ON class.card = book.card;

CREATE INDEX idx_class_card ON class(card);

CREATE INDEX idx_book_card ON book(card);

*使用LEFT JOIN，前面的是驱动表、后面是被驱动表

针对两张表的连接条件涉及的列，索引要创建在被驱动表上，驱动表尽量是小表

如果驱动表上没有where过滤条件
- 当驱动表的连接条件没有索引时，驱动表是全表扫描
- 当针对驱动表的连接条件建立索引时，驱动表依然要进行全索引扫描
- 因此，此时建立在驱动表上的连接条件上的索引是没有太大意义的
如果驱动表上有where过滤条件，那么针对过滤条件创建的索引是有必要的

3、内连接实例

EXPLAIN SELECT * FROM class INNER JOIN book ON class.card = book.card;

CREATE INDEX idx_class_card ON class(card);

CREATE INDEX idx_book_card ON book(card);

*使用INNER JOIN，驱动表、被驱动表不固定，mysql选择

MySQL优化器也会自动选择驱动表，自动选择驱动表的原则是：索引创建在被驱动表上，驱动表是小表。

4、分析4种查询sql（mysql5）

#1 NO3

EXPLAIN SELECT ab.name,c.`name` ceoname FROM

(SELECT a.`name`,b.`CEO` FROM emp a

LEFT JOIN dept b ON a.`deptId`=b.`id`)ab

LEFT JOIN emp c ON ab.ceo=c.`id`;



#2 NO4

EXPLAIN SELECT c.name,ab.name ceoname FROM emp c LEFT JOIN

(SELECT a.`name`,b.`id` FROM emp a

INNER JOIN dept b ON b.`CEO` = a.`id`)ab

ON c.`deptId`= ab.id;



#3  NO1

 EXPLAIN SELECT a.`name`,c.`name` ceoname FROM emp a

LEFT JOIN dept b  ON a.`deptId`= b.id

LEFT JOIN emp c ON b.`CEO`= c.`id`;



#4  NO2

EXPLAIN SELECT a.`name`,(SELECT c.name FROM emp c WHERE c.id =b.`CEO`)ceoname

 FROM emp a

LEFT JOIN dept b ON a.`deptId`=b.`id`;

5、总结

保证被驱动表的JOIN字段已经创建了索引
需要JOIN 的字段，数据类型保持绝对一致。
LEFT JOIN 时，选择小表作为驱动表，大表作为被驱动表。减少外层循环的次数。
INNER JOIN 时，MySQL会自动将小结果集的表选为驱动表。选择相信MySQL优化策略。
能够直接多表关联的尽量直接关联，不用子查询。(减少查询的趟数)
衍生表建不了索引（MySQL5.5）

其他优化

1、子查询优化

（1）获取非掌门人成员

#获取非掌门人成员

CALL proc_drop_index("atguigudb","emp");

CALL proc_drop_index("atguigudb","dept");

SELECT * FROM t_emp a WHERE a.id NOT IN

(SELECT b.ceo FROM t_dept b WHERE b.ceo IS NOT NULL);

EXPLAIN SELECT * FROM emp a WHERE a.id NOT IN

(SELECT b.ceo FROM dept b WHERE b.ceo IS NOT NULL);

#子查询优化NOT IN

EXPLAIN SELECT * FROM emp a LEFT JOIN dept b ON a.id = b.ceo

WHERE b.id IS NULL;

（2）结论

尽量不要使用NOT IN 或者 NOT EXISTS，用LEFT JOIN xxx ON xx = xx WHERE xx IS NULL替代

2、排序优化

（1）实例

CALL proc_drop_index("atguigudb","emp");

CALL proc_drop_index("atguigudb","dept");

CREATE INDEX idx_age_deptid_name ON emp (age,deptid,`name`);

#无过滤，不索引

EXPLAIN SELECT * FROM emp ORDER BY age,deptid;

EXPLAIN SELECT * FROM emp ORDER BY age,deptid LIMIT 10;

EXPLAIN SELECT * FROM emp WHERE age=45 ORDER BY deptid;

#顺序错，不索引

EXPLAIN SELECT * FROM emp WHERE age=45 ORDER BY deptid, `name`;

EXPLAIN SELECT * FROM emp WHERE age=45 ORDER BY deptid, empno;

CREATE INDEX idx_age_deptid_empno ON emp (age,deptid,`empno`);

EXPLAIN SELECT * FROM emp WHERE age=45 ORDER BY `name`, deptid;

EXPLAIN SELECT * FROM emp WHERE deptid=45 ORDER BY age;

#方向反，不索引

EXPLAIN SELECT * FROM emp WHERE age=45 ORDER BY deptid DESC, `name` DESC;

EXPLAIN SELECT * FROM emp WHERE age=45 ORDER BY deptid ASC, `name` DESC;

总结

无过滤，不索引

顺序错，不索引

方向反，不索引

3、mysql索引选择

EXPLAIN SELECT * FROM emp WHERE age =30 AND empno <101000 ORDER BY `name`;

CREATE INDEX idx_age_empno ON emp (age,`empno`);

CREATE INDEX idx_age_name ON emp (age,NAME);

*当【范围条件】和【group by 或者 order by】的字段出现二选一时，优先观察条件字段的过滤数量，如果过滤的数据足够多，而需要排序的数据并不多时，优先把索引放在范围字段上。反之，亦然。

也可以将选择权交给MySQL：索引同时存在，mysql自动选择最优的方案：（对于这个例子，mysql选择idx_age_empno），但是，随着数据量的变化，选择的索引也会随之变化的。

4、双路排序和单路排序

（1）双路排序（慢）

取一批数据，要对磁盘进行两次扫描。众所周知，IO是很耗时的，所以在mysql4.1之后，出现了第二种改进的算法，就是单路排序。

（2）单路排序（快）

它的效率更快一些，因为只读取一次磁盘，避免了第二次读取数据。并且把随机IO变成了顺序IO。但是它会使用更多的空间，因为它把每一行都保存在内存中了。

5、分组优化

group by 使用索引的原则几乎跟order by一致。但是group by 即使没有过滤条件用到索引，也可以直接使用索引（Order By 必须有过滤条件才能使用上索引）
包含了order by、group by、distinct这些查询的语句，where条件过滤出来的结果集请保持在1000行以内，否则SQL会很慢。

6、覆盖索引优化

总结

禁止使用select *，禁止查询与业务无关字段
尽量利用覆盖索引

慢查询日志

1、如何对系统查询慢做索引优化

（1）找运维人员开启生产数据库慢查询日志

（2）等待1-2周时间，积累慢查询日志

（3）借助工具获取慢查询次数最多和查询时间最长的几个sql进行优化

（4）在生产数据库，使用EXPLAIN进行sql分析，找到瓶颈，创建索引优化

（5）关闭慢查询日志。

2、是什么

一种日志记录，查看哪些SQL超出了我们的最大忍耐时间值。

3、使用

（1）开启slow_query_log

SET GLOBAL slow_query_log=1;

SHOW VARIABLES LIKE '%slow_query_log%';

（2）修改long_query_time阈值

SHOW VARIABLES LIKE '%long_query_time%'; -- 查看值：默认10秒

SET GLOBAL long_query_time=0.1; -- 设置一个比较短的时间，便于测试

（3）运行sql

（4）查看慢查询日志

（5）使用工具分析慢查询日志

-- 查看mysqldumpslow的帮助信息

mysqldumpslow --help

-- 工作常用参考

-- 1.得到返回记录集最多的10个SQL

mysqldumpslow -s r -t 10 /var/lib/mysql/atguigu-slow.log

-- 2.得到访问次数最多的10个SQL

mysqldumpslow -s c -t 10 /var/lib/mysql/atguigu-slow.log

-- 3.得到按照时间排序的前10条里面含有左连接的查询语句

mysqldumpslow -s t -t 10 -g "left join" /var/lib/mysql/atguigu-slow.log

-- 4.另外建议在使用这些命令时结合 | 和more 使用，否则语句过多有可能出现爆屏情况

mysqldumpslow -s r -t 10 /var/lib/mysql/atguigu-slow.log | more

1、单表索引失效案例

2、关联查询优化

3、其他优化

4、慢查询日志

5、视图

6、高性能架构模式