同一列的多个过滤索引

Question

根据不同条件为同一列创建多个过滤索引是否合理？例如，假设以下内容：

CREATE TABLE [dbo].[Skills](
    [UserId] INT NOT NULL PRIMARY KEY, 
    [SkillTitle] NVARCHAR(50) NOT NULL, 
    [SkillLevel] INT NOT NULL CHECK ([SkillLevel] BETWEEN 0 AND 10) 
) 
GO 

CREATE INDEX [IX_Skill_Low] ON [dbo].[Skill] (SkillTitle,SkillLevel) WHERE SkillLevel BETWEEN 3 AND 0 
GO 

CREATE INDEX [IX_Skill_Moderate] ON [dbo].[Skill] (SkillTitle,SkillLevel) WHERE SkillLevel BETWEEN 5 AND 3 
GO 

CREATE INDEX [IX_Skill_Good] ON [dbo].[Skill] (SkillTitle,SkillLevel) WHERE SkillLevel BETWEEN 7 AND 5 
GO 

CREATE INDEX [IX_Skill_High] ON [dbo].[Skill] (SkillTitle,SkillLevel) WHERE SkillLevel > 7 

或者创建单个非过滤索引会更好吗？

如果进行以下查询，会发生什么SELECT * FROM Skills WHERE SkillLevel BETWEEN 7 AND 1？

Answer 1

除非您有数百万记录用于每种技能等级的变化。

即，
在5至7

在5和3
我不会建议，我建议建立像下面

skilllevel between 1 and 10 

是怎样的一个单一过滤指数如果进行以下查询，则会发生SELECT * FROM技能WHERE SkillLevel BETWEEN 7 AND 1？

如果如果你select *包含许多列以外，其在指数没有specfied，SQL可能

• 选择使用该索引和做基表查找的列的其余
• 做一个全表扫描

我会创造一个下方覆盖指数，但是这取决于你的数据

create index nci_test on table(skilllevel) 
include(SkillTitle) 

我做了一些测试，以了解更多，你可以找到下面的测试脚本

CREATE TABLE [dbo].[Skills](
    [UserId] INT NOT NULL identity(1,1) PRIMARY KEY, 
    [SkillTitle] NVARCHAR(50) NOT NULL, 
    [SkillLevel] INT NOT NULL CHECK ([SkillLevel] BETWEEN 0 AND 10) 
) 
GO 

insert into dbo.Skills 
(SkillTitle,SkillLevel) 
select cast(NEWID() as varchar(50)), 
case when rand()*10<0 then 1 else rand()*10 end 
go 100000 

现在让我们运行一些查询

create index nci_1and5 on dbo.skills(skilllevel) 
include(skilltitle) 
where skilllevel >= 1 and skilllevel <=5 

create index nci_5and10 on dbo.skills(skilllevel) 
include(skilltitle) 
where skilllevel >5 and skilllevel <=10 

select * from skills where skilllevel 
between 1 and 4--uses 1 to 5 index 

select * from skills where skilllevel 
between 6 and 10--uses 6 to 10 index 

select * from skills where skilllevel 
between 1 and 7--uses scan 

但是我不推荐这些指标，即使你有百万行之间的技能levels.I将建议分区