简单谈谈MySQL的loose index scan
众所周知,innodb采用iot(index organization table)即所谓的索引组织表,而叶子节点也就存放了所有的数据,这就意味着,数据总是按照某种顺序存储的。所以问题来了,如果是这样一个语句,执行起来应该是怎么样的呢?语句如下:
select count(distinct a) from table1;
列a上有一个索引,那么按照简单的想法来讲,如何扫描呢?很简单,一条一条的扫描,这样一来,其实做了一次索引全扫描,效率很差。这种扫描方式会扫描到很多很多的重复的索引,这样说的话优化的办法也是很容易想到的:跳过重复的索引就可以了。于是网上能搜到这样的一个优化的办法:
select count(*) from (select distinct a from table1) t;
从已经搜索到的资料看,这样的执行计划中的extra就从using index变成了using index for group-by。
但是,但是,但是,好在我们现在已经没有使用5.1的版本了,大家基本上都是5.5以上了,这些现代版本,已经实现了loose index scan:
很好很好,就不需要再用这种奇技淫巧去优化sql了。
文档里关于group by这里写的有点意思,说是最大众化的办法就是进行全表扫描并且创建一个临时表,这样执行计划就会难看的要命了,肯定有all和using temporary table了。
5.0之后group by在特定条件下可能使用到loose index scan,
create table log_table ( id int not null primary key, log_machine varchar(20) not null, log_time datetime not null ) engine=innodb default charset=utf8; create index ix_log_machine_time on log_table (log_machine, log_time);
1
select max(log_time) from log_table; select max(log_time) from log_table where log_machine in ('machine 1');
这两条sql都只需一次index seek便可返回,源于索引的有序排序,优化器意识到min/max位于最左/右块,从而避免范围扫描;
extra显示select tables optimized away ;
2
执行计划type 为range(extra显示using where; using index),即执行索引范围扫描,先读取所有满足log_machine约束的记录,然后对其遍历找出max value;
改进
这满足group by选择loose index scan的要求,执行计划的extra显示using index for group-by,执行效果等值于
select max(log_time) from log_table where log_machine in (‘machine 1') union select max(log_time) from log_table where log_machine in (‘machine 2') …..
即对每个log_machine执行loose index scan,rows从原来的82636下降为16(该表总共1,000,000条记录)。
group by何时使用loose index scan?
适用条件:
1 针对单表操作
2 group by使用索引的最左前缀列
3 只支持聚集函数min()/max()
4 where条件出现的列必须为=constant操作 , 没出现在group by中的索引列必须使用constant
5 不支持前缀索引,即部分列索引 ,如index(c1(10))
执行计划的extra应该显示using index for group-by
假定表t1有个索引idx(c1,c2,c3)
select c1, c2 from t1 group by c1, c2; select distinct c1, c2 from t1; select c1, min(c2) from t1 group by c1; select c1, c2 from t1 where c1 < const group by c1, c2; select max(c3), min(c3), c1, c2 from t1 where c2 > const group by c1, c2; select c2 from t1 where c1 < const group by c1, c2; select c1, c2 from t1 where c3 = const group by c1, c2 select c1, c3 from t1 group by c1, c2;--无法使用松散索引
而select c1, c3 from t1 where c3= const group by c1, c2;则可以
紧凑索引扫描tight index scan
group by在无法使用loose index scan,还可以选择tight,若两者都不可选,则只能借助临时表;
扫描索引时,须读取所有满足条件的索引键,要么是全索引扫描,要么是范围索引扫描;
group by的索引列不连续;或者不是从最左前缀开始,但是where条件里出现最左列;
select c1, c2, c3 from t1 where c2 = 'a' group by c1, c3; select c1, c2, c3 from t1 where c1 = 'a' group by c2, c3;
5.6的改进
事实上,5.6的index condition push down可以弥补loose index scan缺失带来的性能损失。
key(age,zip)
mysql> explain select name from people where age between 18 and 20 and zip in (12345,12346, 12347); +----+-------------+--------+-------+---------------+------+---------+------+-------+-------------+ | id | select_type | table | type | possible_keys | key | key_len | ref | rows | extra | +----+-------------+--------+-------+---------------+------+---------+------+-------+-------------+ | 1 | simple | people | range | age | age | 4 | null | 90556 | using where | +----+-------------+--------+-------+---------------+------+---------+------+-------+-------------+ 1 row in set (0.01 sec)
根据key_len=4可以推测出sql只用到索引的第一列,即先通过索引查出满足age (18,20)的行记录,然后从server层筛选出满足zip约束的行;
pre-5.6,对于复合索引,只有当引导列使用"="时才有机会在索引扫描时使用到后面的索引列。
mysql> explain select name from people where age=18 and zip in (12345,12346, 12347); +----+-------------+--------+-------+---------------+------+---------+------+------+-------------+ | id | select_type | table | type | possible_keys | key | key_len | ref | rows | extra | +----+-------------+--------+-------+---------------+------+---------+------+------+-------------+ | 1 | simple | people | range | age | age | 8 | null | 3 | using where | +----+-------------+--------+-------+---------------+------+---------+------+------+-------------+ 1 row in set (0.00 sec)
对比一下查询效率
mysql> select sql_no_cache name from people where age=19 and zip in (12345,12346, 12347); +----------------------------------+ | name | +----------------------------------+ | 888ba838661aff00bbbce114a2a22423 | +----------------------------------+ 1 row in set (0.06 sec) mysql> select sql_no_cache name from people where age between 18 and 22 and zip in (12345,12346, 12347); +----------------------------------+ | name | +----------------------------------+ | ed4481336eb9adca222fd404fa15658e | | 888ba838661aff00bbbce114a2a22423 | +----------------------------------+ 2 rows in set (1 min 56.09 sec)
对于第二条sql,可以使用union改写,
mysql> select name from people where age=18 and zip in (12345,12346, 12347) -> union all -> select name from people where age=19 and zip in (12345,12346, 12347) -> union all -> select name from people where age=20 and zip in (12345,12346, 12347) -> union all -> select name from people where age=21 and zip in (12345,12346, 12347) -> union all -> select name from people where age=22 and zip in (12345,12346, 12347);
而mysql5.6引入了index condition pushdown,从优化器层面解决了此类问题。
上一篇: EJB中的依赖管理
下一篇: 010_JS中常用的几个事件的学习