Question

这是我的疑问：

SELECT * 
FROM 
    (SELECT 
        ROW_NUMBER() OVER (ORDER BY NAME asc) peta_rn, 
        peta_query.* 
     FROM 
         (SELECT 
              BOOK, PAGETRIMMED, NAME, TYPE, PDF 
         FROM 
              CCWiseDocumentNames2 cdn
         INNER JOIN 
              CCWiseInstr2 cwi ON cwi.ID = cdn.ID) as peta_query) peta_paged 
WHERE 
    peta_rn > 1331900 AND peta_rn <= 1331950

目前此查询大约需要4秒才能获得结果。有没有办法让它低于1秒？

已在cwi.ID和cdn.ID上创建了索引。以下是来自sql server的实际执行计划：

enter image description here

任何帮助都会有用。

这是表结构：

    /****** Object:  Table [dbo].[CCWiseInstr2]    Script Date: 9/17/2013 3:54:27 AM ******/
    SET ANSI_NULLS ON
    GO

    SET QUOTED_IDENTIFIER ON
    GO

    SET ANSI_PADDING ON
    GO

    CREATE TABLE [dbo].[CCWiseInstr2](
        [ID] [int] NULL,
        [BK_PG] [varchar](50) NULL,
        [DATE] [datetime] NULL,
        [ITYPE] [varchar](50) NULL,
        [BOOK] [int] NULL,
        [PAGE] [varchar](50) NULL,
        [NOBP] [varchar](50) NULL,
        [DESC] [varchar](240) NULL,
        [TIF] [varchar](50) NULL,
        [INDEXNAME] [varchar](50) NULL,
        [CONFIRM] [varchar](50) NULL,
        [PDF] [varchar](50) NULL,
        [PAGETRIMMED] [varchar](10) NULL
    ) ON [PRIMARY]

    GO

    SET ANSI_PADDING OFF
    GO

    /****** Object:  Index [IX_CCWiseInstr2_ID]    Script Date: 9/17/2013 3:54:27 AM ******/
    CREATE NONCLUSTERED INDEX [IX_CCWiseInstr2_ID] ON [dbo].[CCWiseInstr2]
    (
        [ID] ASC
    )WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
    GO



/****** Object:  Table [dbo].[CCWiseDocumentNames2]    Script Date: 9/17/2013 3:54:18 AM ******/
SET ANSI_NULLS ON
GO

SET QUOTED_IDENTIFIER ON
GO

SET ANSI_PADDING ON
GO

CREATE TABLE [dbo].[CCWiseDocumentNames2](
    [ID] [int] NULL,
    [BK_PG] [varchar](50) NULL,
    [NAME] [varchar](100) NULL,
    [OTHERNAM] [varchar](100) NULL,
    [TYPE] [varchar](50) NULL,
    [INDEXNAME] [varchar](50) NULL
) ON [PRIMARY]

GO

SET ANSI_PADDING OFF
GO

/****** Object:  Index [IX_CCWiseDocumentNames2_ID]    Script Date: 9/17/2013 3:54:18 AM ******/
CREATE NONCLUSTERED INDEX [IX_CCWiseDocumentNames2_ID] ON [dbo].[CCWiseDocumentNames2]
(
    [ID] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
GO

SET ANSI_PADDING ON

GO

/****** Object:  Index [IX_CCWiseDocumentNames2_NAME]    Script Date: 9/17/2013 3:54:18 AM ******/
CREATE NONCLUSTERED INDEX [IX_CCWiseDocumentNames2_NAME] ON [dbo].[CCWiseDocumentNames2]
(
    [NAME] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
GO

Answer 1

您不需要拥有PK或Identity，因此您仍然可以在ID列上创建聚簇索引。它只会重复值，你应该担心的唯一问题是INSERT性能，如果你没有附加ID，而是添加到中间。

为什么在外部查询中使用row_number（）？我认为在单选中可以实现相同的结果（也许你必须改变排名功能并使用分区）。顺便说一下，如果您的内部查询没有返回唯一的NAME，并且您在没有分区的情况下使用row_number，那么peta_rn可能会返回误导性值（与许多不同的peta_rn同名）。我只是在猜测，因为我不知道你到底想要达到的目的。

使用聚集索引，你将它带到1s以下没问题。

Answer 2

我认为问题是两个表都需要完全匹配，然后在where子句抛弃垃圾之前按名称排序。

我不确定这会有所帮助，但值得一试 - 尝试将名称添加到索引中：

CREATE NONCLUSTERED INDEX [IX_CCWiseDocumentNames2_ID] ON [dbo].[CCWiseDocumentNames2]
(
    [ID] ASC,
    [Name] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]

如何加快Row_Number查询速度？

2 个答案: