Question

已在dw_assesment_details和dw_assesment_details_id表

上创建了聚簇索引

/* 6 minutes */
CREATE CLUSTERED INDEX [Ix_DW_ASSESSMENT_DETAILS_qid_QNO_TmpverName] 
ON [dbo].[DW_ASSESSMENT_DETAILS_QUESTION_ID]
(
    [TEMPLATENAME] ASC,
    [TEMPLATEVERSION] ASC,
    [QUESTION_NO] 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)
GO

/* 9 minutes */
CREATE CLUSTERED INDEX [Ix_DW_ASSESSMENT_DETAILS_QNO_TmpverName] 
ON [dbo].[DW_ASSESSMENT_DETAILS]
(
    [TEMPLATENAME] ASC,
    [TEMPLATEVERSION] ASC,
    [QUESTION_NO] 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)

SELECT 
    [GETQUESTIONID], 
    dw.[TEMPLATENAME], dw.[TEMPLATEVERSION],
    dw.[QUESTION_NO]
FROM 
    [QIS_DW].[dbo].[DW_ASSESSMENT_DETAILS] dw 
INNER JOIN
    [QIS_DW].[dbo].[DW_ASSESSMENT_DETAILS_QUESTION_ID] id ON dw.TEMPLATENAME = id.TEMPLATENAME 
                                                          AND dw.TEMPLATEVERSION = id.TEMPLATEVERSION 
                                                          AND dw.QUESTION_NO = id.QUESTION_NO

但是上面的select查询是使用索引扫描 - 而不是索引搜索。如何使用索引寻求？

性能调优专家的任何建议？

Answer 1

如何使用索引查找？

如果您想尝试智能SQL Server优化器，可以指定LOOP JOIN查询提示。我希望MERGE JOIN的计划在很多行上表现得更好。

请注意，存储引擎还可能在大型扫描期间执行异步预读，以将数据预取到内存中，以便查询可以使用它，而无需等待从存储中读取。对于返回少量行的搜索，不会发生预读。

尝试以下查询，看看您的环境是否属于这种情况。

SET STATISTICS IO ON;
SET STATISTICS TIME ON;
GO

SELECT 
    [GETQUESTIONID], 
    dw.[TEMPLATENAME], dw.[TEMPLATEVERSION],
    dw.[QUESTION_NO]
FROM 
    [QIS_DW].[dbo].[DW_ASSESSMENT_DETAILS] dw 
INNER JOIN
    [QIS_DW].[dbo].[DW_ASSESSMENT_DETAILS_QUESTION_ID] id ON dw.TEMPLATENAME = id.TEMPLATENAME 
                                                          AND dw.TEMPLATEVERSION = id.TEMPLATEVERSION 
                                                          AND dw.QUESTION_NO = id.QUESTION_NO;
GO

SELECT 
    [GETQUESTIONID], 
    dw.[TEMPLATENAME], dw.[TEMPLATEVERSION],
    dw.[QUESTION_NO]
FROM 
    [QIS_DW].[dbo].[DW_ASSESSMENT_DETAILS] dw 
INNER JOIN
    [QIS_DW].[dbo].[DW_ASSESSMENT_DETAILS_QUESTION_ID] id ON dw.TEMPLATENAME = id.TEMPLATENAME 
                                                          AND dw.TEMPLATEVERSION = id.TEMPLATEVERSION 
                                                          AND dw.QUESTION_NO = id.QUESTION_NO
OPTION(LOOP JOIN);
GO

Answer 2

在SQL Server中，Clustered或Non-Clustered索引表搜索更改索引扫描，直到我们没有在Where子句中添加条件。

Where子句会使Index寻求执行计划。如果没有Where子句，它将作为索引扫描来搜索整个表数据。这就是索引扫描总是很慢的原因。

为什么这是索引扫描而不是索引搜索

2 个答案: