Question

假设我有两张表Company（差不多60K记录）和Position（差不多600K记录）

Company表：

CompanyID     INT                --PRIMARY KEY
CompanyName   NVARCHAR(100)      
CompanyType   INT               --Just can be (1,2,3,4,5,6)

Position表：

PositionID       INT             --Primary key
PositionName     NVARCHAR(100)   
CompanyID        INT             --FK point to Company Table
WorkExperience   INT             --Just can be (1,2,3,4,5,6,7,8) 
WorkType         INT             --Just can be (1,2) 
CreateTime       datetime
UpdateTime       datetime

我在NONCLUSTERED INDEX表格上创建了Company：

CREATE NONCLUSTERED INDEX [IX_1] ON [dbo].[Company] 
(
    [CompanyKind] ASC
)
INCLUDE ( [CompanyName]) ON [PRIMARY]
GO

我在NONCLUSTERED INDICES表上创建了两个Position：

CREATE NONCLUSTERED INDEX [IX_6] ON [dbo].[Position] 
(
    [CompanyID] ASC
)ON [PRIMARY]

CREATE NONCLUSTERED INDEX [IX_8] ON [dbo].[Position] 
(
    [UpdateTime] ASC
) ON [PRIMARY]

我的分页存储过程如下所示：

ALTER PROC [dbo].[spIndexJobList]
    @KeyWord NVARCHAR(50) ,
    @WorkExperience INT ,
    @WorkType INT ,
    @CompanyType INT ,
    @PageSize INT ,
    @PageNumber INT 
    --@RowCount INT OUTPUT
AS 
    DECLARE @RowStart INT
    DECLARE @RowEnd INT
    DECLARE @SQL NVARCHAR(4000)
    DECLARE @ParamDefinition NVARCHAR(2000) 

    SET @SQL = N'SELECT C.CompanyID,C.CompanyName,P.PositionName,P.PositionID,P.UpdateTime, Row_number() OVER (ORDER BY P.UpdateTime DESC) AS RowNumber FROM Company C INNER JOIN Position P ON C.CompanyID=P.CompanyID WHERE 1=1 '
    IF @KeyWord!=''
      SET @SQL = @SQL + ' AND PositionName LIKE @KeyWord'
    IF @WorkExperience !=0 
        SET @SQL = @SQL + ' AND P.WorkExperience=@WorkExperience'
    IF @CompanyType != 0 
        SET @SQL = @SQL + ' AND C.CompanyType=@CompanyType'
    IF @WorkType !=0
        SET @SQL = @SQL + ' AND P.WorkType=@WorkType'
    SET @ParamDefinition = ' @KeyWord    NVarchar(50),
                             @WorkExperience   INT,
                             @WorkType       INT,
                             @CompanyType       INT,
                             @PageSize   INT,
                             @PageNumber INT'
    IF @PageNumber > 0 
        BEGIN
            SET @PageNumber = @PageNumber - 1
            SET @RowStart = @PageSize * @PageNumber + 1 ;
            SET @RowEnd = @RowStart + @PageSize - 1 ;
            SET @SQL = '
        WITH AllJobs
             AS (' + @SQL
                + ')

   SELECT *,(SELECT Count(RowNumber)  FROM   AllJobs) AS TotalRows FROM   AllJobs  WHERE  RowNumber >='
                + STR(@RowStart) + '  AND RowNumber <= ' + STR(@RowEnd) + ''

            EXECUTE sp_Executesql @SQL, @ParamDefinition,
                @KeyWord, @WorkExperience,@WorkType, 
                @CompanyType, @PageSize, @PageNumber

        END

我的电话声明是：

SET STATISTICS IO ON
DECLARE @return_value int
EXEC    @return_value = [dbo].[spIndexJobList]
        @KeyWord='',
        @WorkExperience = 3,
        @CompanyType = 2,
        @WorkType =1,
        @PageSize = 30,
        @PageNumber =2000

SELECT  'Return Value' = @return_value
GO
SET STATISTICS IO OFF

// --------------------------------------------- --------------------------

(30 row(s) affected)
Table 'Company'. Scan count 3, logical reads 632, physical reads 0
Table 'Position'. Scan count 3, logical reads 4865, physical reads 0
Table 'Worktable'. Scan count 0, logical reads 0, physical reads 0
Table 'Worktable'. Scan count 0, logical reads 0, physical reads 0

// --------------------------------------------- --------------------------

enter image description here

执行计划提示缺少索引（WorkExperience，WorkType）。但是在创建索引（WorkExperience，WorkType）之后查询速度较慢。

任何给我一些建议的人都会非常感激。抱歉我的英文不好！

Answer 1

<强> DDL：

公司表：

CompanyID     INT NOT NULL PK
CompanyName   NVARCHAR(100) NOT NULL     
CompanyType   TINYINT NOT NULL

职位表：

PositionID       INT NOT NULL PK
PositionName     NVARCHAR(100) NOT NULL   
CompanyID        INT NOT NULL
WorkExperience   TINYINT NOT NULL
WorkType         TINYINT NOT NULL
CreateTime       SMALLDATETIME
UpdateTime       SMALLDATETIME

<强>索引：

CREATE NONCLUSTERED INDEX IX_Position
     ON Position (WorkExperience, WorkType, PositionName)
     INCLUDE (UpdateTime)

<强> SP：

ALTER PROC [dbo].[spIndexJobList]

      @KeyWord NVARCHAR(50) 
    , @WorkExperience TINYINT 
    , @WorkType TINYINT 
    , @CompanyType INT 
    , @PageSize INT 
    , @PageNumber INT 

AS BEGIN

     SET NOCOUNT ON;

     IF @PageNumber > 0 BEGIN

          DECLARE 
                 @RowStart INT
               , @RowEnd INT
               , @SQL NVARCHAR(4000)
               , @ParamDefinition NVARCHAR(500) 

          SELECT @SQL = N'
               SELECT 
                      c.CompanyID
                    , c.CompanyName
                    , p.PositionName
                    , p.PositionID
                    , p.UpdateTime
                    , RowNumber = ROW_NUMBER() OVER (ORDER BY p.UpdateTime DESC)  
               FROM dbo.Company c 
               JOIN dbo.Position p ON c.CompanyID = p.CompanyID 
               WHERE 1=1 '
               + CASE WHEN @KeyWord != '' THEN ' AND PositionName LIKE @KeyWord' ELSE '' END
               + CASE WHEN @WorkExperience != 0 THEN ' AND p.WorkExperience = @WorkExperience' ELSE '' END
               + CASE WHEN @CompanyType != 0 THEN ' AND c.CompanyType = @CompanyType' ELSE '' END
               + CASE WHEN @WorkType != 0 THEN ' AND p.WorkType = @WorkType' ELSE '' END

          SET @ParamDefinition = '@KeyWord NVARCHAR(50),
                                  @WorkExperience INT,
                                  @WorkType INT,
                                  @CompanyType INT,
                                  @PageSize INT,
                                  @PageNumber INT'

          SELECT 
                 @PageNumber = @PageNumber - 1
               , @RowStart = @PageSize * @PageNumber + 1
               , @RowEnd = @RowStart + @PageSize - 1
               , @SQL = '
          WITH AllJobs AS (' + @SQL + ')
          SELECT *
          FROM AllJobs a
          CROSS JOIN (
               SELECT TotalRows = Count(RowNumber)  
               FROM AllJobs
          ) t
          WHERE a.RowNumber BETWEEN ' + STR(@RowStart) + ' AND ' + STR(@RowEnd)

          EXEC sys.sp_executesql 
               @SQL, 
               @ParamDefinition,
               @KeyWord, 
               @WorkExperience,
               @WorkType, 
               @CompanyType, 
               @PageSize, 
               @PageNumber

     END

END

<强>手册

Bad habits to kick : choosing the wrong data type

Answer 2

我会在Position上创建一个非聚集索引，如下所示：

CREATE NONCLUSTERED INDEX IX_Position_WorkExperienceWorkTypePositionName 
ON Position (WorkExperience, WorkType, PositionName)
INCLUDE (PositionID, UpdateTime)

Answer 3

另外，我会在公司上创建另一个非聚集索引，如下所示：

CREATE NONCLUSTERED INDEX IX_Company_CompanyId_Type
ON Company(CompanyId,CompanyType)

因为where子句包含“C.CompanyType=@CompanyType”，没有这个索引，这个条件导致了表扫描公司。有了这个索引，它就是索引搜索。

Answer 4

应该对Company.CompanyID和Position.CompanyID建立索引，您也可以利用CLUSTER对这两列进行索引，以便对这些行进行物理排序，使其与索引同步。这应该可以显着提高性能。

如何提高这个sql查询性能？

4 个答案: