How do I speed up this Row_Number query?

This is my query:

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

Currently this query takes about 4 seconds to get the results. Is there any way to bring it under 1 second?

Index is already created on cwi.ID and cdn.ID. Below is the actual execution plan from sql server:

Any help would be useful.

This is the table structure:

    /****** 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

回答1:

I think the problem is that both tables need to be matched completely and then sorted on name before the where clause can throw out the trash.

I'm not sure this will help, but it's worth a shot - try adding the name to the index:

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]

回答2:

You don't need to have PK or Identity, so you still can create clustered index on ID column. It alows repeating values, and the only thing you should be worrying about is INSERT performance if you're the IDs are not appended, but added in the middle.

Why are you using row_number() in outer query? I think the same result can be achieved in single select (maybe you'll have to change the ranking function and use partitioning). Btw if your inner query does not return unique NAME and you're using row_number without partitioning, then peta_rn may return misleading values (same name with many different peta_rn). I'm just guessing because I don't know what exactly are you trying to achieve.

Go with the clustered index and you'll bring it under 1s no problem.