如何知道SQL SERVER索引的使用情况
如何提高查询语句的效率？一个有效的手段是创建索引。那么怎么知道你创建的索引的使用情况呢？showplan_text是个不错的办法。下面贴点代码说一说吧。
CREATE
TABLE
[
dbo
]
.
[
MyTable
]
(2
[
ID
]
[
int
]
IDENTITY
(
1
,
1
)
NOT
NULL
,3
[
Column1
]
[
int
]
NOT
NULL
,4
[
Column2
]
[
int
]
NOT
NULL
,5
[
Column3
]
[
int
]
NOT
NULL
,6
CONSTRAINT
[
PK_MyTable
]
PRIMARY
KEY
CLUSTERED
7
(8
[
ID
]
ASC
9
)
WITH
(IGNORE_DUP_KEY
=
OFF
)
ON
[
PRIMARY
]
10
)
ON
[
PRIMARY
]
11
go
12
CREATE
NONCLUSTERED
INDEX
[
IX_MyTable
]
ON
[
dbo
]
.
[
MyTable
]
13
(14
[
Column1
]
ASC
,15
[
Column2
]
ASC
,16
[
Column3
]
ASC
17
)
WITH
(SORT_IN_TEMPDB
=
OFF
, DROP_EXISTING
=
OFF
, IGNORE_DUP_KEY
=
OFF
, ONLINE
=
OFF
)
ON
[
PRIMARY
]
18
19
go
20
21
set
showplan_text
on
22
go
23
24
select
*
from
MyTable25
where
26
Column1
=
1
and
Column2
=
1
and
Column3
=
1
27
go
28
29
set
showplan_text
off
30
go
31
32
drop
table
MyTable33
go
表上所建的索引为
Column1 ASC
Column2 ASC
Column3 ASC
1.where 语句条件同索引
select * from MyTable where Column1=1 and Column2=1 and Column3=1
结果为
|--Index Seek(OBJECT:([Test].[dbo].[MyTable].[IX_MyTable]), SEEK:([Test].[dbo].[MyTable].[Column1]=CONVERT_IMPLICIT(int,[@1],0) AND [Test].[dbo].[MyTable].[Column2]=CONVERT_IMPLICIT(int,[@2],0) AND [Test].[dbo].[MyTable].[Column3]=CONVERT_IMPLICIT(int,[@3],0)) ORDERED FORWARD)
没有问题，索引完全发挥了作用。
2.where 语句条件与索引顺序不同
select * from MyTable where Column3=1 and Column2=1 and Column1=1
|--Index Seek(OBJECT:([Test].[dbo].[MyTable].[IX_MyTable]), SEEK:([Test].[dbo].[MyTable].[Column1]=CONVERT_IMPLICIT(int,[@2],0) AND [Test].[dbo].[MyTable].[Column2]=CONVERT_IMPLICIT(int,[@3],0) AND [Test].[dbo].[MyTable].[Column3]=CONVERT_IMPLICIT(int,[@1],0)) ORDERED FORWARD)
实际执行经过优化，此时索引最佳。
3.where 语句条件比索引列少
select * from MyTable where Column1=1 and Column2=1
|--Index Seek(OBJECT:([Test].[dbo].[MyTable].[IX_MyTable]), SEEK:([Test].[dbo].[MyTable].[Column1]=CONVERT_IMPLICIT(int,[@1],0) AND [Test].[dbo].[MyTable].[Column2]=CONVERT_IMPLICIT(int,[@2],0)) ORDERED FORWARD)
索引也同样发挥了作用。
4.where 语句条件比索引列少，起始列相同
select * from MyTable where Column1=1 and Column3=1
|--Index Seek(OBJECT:([Test].[dbo].[MyTable].[IX_MyTable]), SEEK:([Test].[dbo].[MyTable].[Column1]=CONVERT_IMPLICIT(int,[@1],0)), WHERE:([Test].[dbo].[MyTable].[Column3]=CONVERT_IMPLICIT(int,[@2],0)) ORDERED FORWARD)
5.where 语句与索引顺序不一致，缺少第一列
select * from MyTable where Column3=1 and Column2=1
|--Index Scan(OBJECT:([Test].[dbo].[MyTable].[IX_MyTable]), WHERE:([Test].[dbo].[MyTable].[Column3]=CONVERT_IMPLICIT(int,[@1],0) AND [Test].[dbo].[MyTable].[Column2]=CONVERT_IMPLICIT(int,[@2],0)))
Index Scan，而不是Index Seek，后者性能优于前者。
6.where 语句与索引顺序不一致，有第一列
select * from MyTable where Column3=1 and Column1=1
|--Index Seek(OBJECT:([Test].[dbo].[MyTable].[IX_MyTable]), SEEK:([Test].[dbo].[MyTable].[Column1]=CONVERT_IMPLICIT(int,[@2],0)), WHERE:([Test].[dbo].[MyTable].[Column3]=CONVERT_IMPLICIT(int,[@1],0)) ORDERED FORWARD)
结论：
与索引列完全一致的where条件（and连接，使用=），此时的sql语句最为高效。
实际执行的查询经过优化，最大限度地使用索引。
有索引第一列时，将是Index Seek，否则（有索引其他列）是Index Scan。
showplan_text显示的信息我还得再学习学习，如果谁能提供点资料真是感激不尽。欢迎指点！