Question

我有一个慢查询，其中包含10个以上的左连接和10个类似的运算符。见吼叫。

SystemContact约有。 10K的数据，另外两个有超过100k的数据。查询看起来没有优化，这是因为它是由用于搜索SystemContact的程序生成的。

我遇到的问题是，查询在我们的系统上运行速度非常慢，至少需要一分钟才能运行。我们已经完成了更新统计信息和对表进行碎片整理，但它似乎并没有提高性能。

最初我认为它可能与当前使用nvarchar的字符串DataType有关，但是当我将所有那些带有数据的表加载到同一服务器上的tempdb时，查询运行大约2秒。

我想知道造成这种缓慢的原因是什么原因？

SET TRANSACTION ISOLATION LEVEL READ COMMITTED;

    select
    Id, Context1, Context2, Context3, LastModifiedBy

    from SystemContact
    left join SystemField as SF1 on SystemContact.Id = SF1.ContactId and SF1.FieldId = 1111
    left join SystemField as SF2 on SystemContact.Id = SF2.ContactId and SF2.FieldId = 5445
    left join SystemField as SF3 on SystemContact.Id = SF3.ContactId and SF3.FieldId = 3423
    left join SystemField as SF4 on SystemContact.Id = SF4.ContactId and SF4.FieldId = 6545
    left join SystemField as SF5 on SystemContact.Id = SF5.ContactId and SF5.FieldId = 5464
    left join SystemFieldText as SFT1 on SystemContact.Id = SFT1.ContactId and SFT1.FieldId = 546
    left join SystemFieldText as SFT2 on SystemContact.Id = SFT2.ContactId and SFT2.FieldId = 7565
    left join SystemFieldText as SFT3 on SystemContact.Id = SFT3.ContactId and SFT3.FieldId = 456
    left join SystemFieldText as SFT4 on SystemContact.Id = SFT4.ContactId and SFT4.FieldId = 457
    left join SystemFieldText as SFT5 on SystemContact.Id = SFT5.ContactId and SFT5.FieldId = 56
    left join SystemFieldText as SFT6 on SystemContact.Id = SFT6.ContactId and SFT6.FieldId = 85

    where
    (SF1.Value like '%john.smith@example.com%') or (SF2.Value like '%john.smith@example.com%') or (SF3.Value like '%john.smith@example.com%') 
    or (SF4.Value like '%john.smith@example.com%') or (SF5.Value like '%john.smith@example.com%') 
    or (SFT2.Value like '%john.smith@example.com%') or (SFT3.Value like '%john.smith@example.com%') or (SFT4.Value like '%john.smith@example.com%')
    or (SFT5.Value like '%john.smith@example.com%') or (SFT6.Value like '%john.smith@example.com%')

编辑：

以下是执行计划的链接：https://goo.gl/gICWir

Answer 1

CREATE NONCLUSTERED INDEX ix
    ON dbo.SystemFieldText (FieldId, ContactId) INCLUDE (value)
GO

SELECT id,
       Context1,
       Context2,
       Context3,
       LastModifiedBy
FROM dbo.SystemContact
WHERE EXISTS(
        SELECT *
        FROM dbo.SystemField SF
        WHERE SystemContact.id = SF.ContactId
            AND SF.value LIKE '%john.smith@example.com%'
            AND SF.FieldId IN (1111, 5445, 3423, 6545, 5464)
    )
    OR EXISTS(
        SELECT *
        FROM dbo.SystemFieldText
        WHERE SystemContact.id = SFT.ContactId
            AND SFT.value LIKE '%john.smith@example.com%'
            AND SFT.FieldId IN (546, 7565, 7565, 456, 457, 56, 85)
    )

SQL Server查询速度很慢，但在tempdb上进行测试时，它的工作速度很快

1 个答案: