如何提高少量更新行的更新速度？

Question

我正在使用 SQL Server 2019，并面临一个问题，即在进行更新时，只需要 20 秒即可更新 162 行。

• 预计执行计划

• 实际执行计划

Update FT SET
    ft.ValueName=avo.name
FROM #package FT 
inner join parts.Nop_PackageAttribute PA with(nolock)
    on PA.PackageID=ft.PackageID
        and PA.[Key]=FT.ZfeatureId 
inner join Nop_AcceptedValuesOption AVO with(nolock)
    ON convert(varchar(20),AVO.AcceptedValuesOptionID)=PA.Value 
where FT.AcceptedValueID is not null

脚本示例

 create table #package
 (
 id int PRIMARY KEY IDENTITY(1,1),
 ZfeatureId INT NULL,
 AcceptedValueID INT NULL,
 PackageID INT NULL,
 ValueName NVARCHAR(2000) default ''
 )

我在 #package 表上的索引

 create nonclustered index IDX_PackageID on #package(PackageID) include (ZfeatureId,AcceptedValueID , ValueName)
 create index acceptedvaluesidpackage_idx on #package(AcceptedValueID)

包属性表

ALTER TABLE [Parts].[Nop_PackageAttribute] ADD  CONSTRAINT [PK_Nop_PackageAttribute] PRIMARY KEY CLUSTERED 
(
    [PackageAttributeID] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [Customer]
GO
CREATE NONCLUSTERED INDEX [_dta_index_Nop_PackageAttribute_8_578153155__K2_K1_K3_4] ON [Parts].[Nop_PackageAttribute]
(
    [PackageID] ASC,
    [PackageAttributeID] ASC,
    [Key] ASC
)
INCLUDE (   [Value]) 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 [Customer]

CREATE NONCLUSTERED INDEX [IDX_Key] ON [Parts].[Nop_PackageAttribute]
(
    [Key] 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 [Customer]


CREATE NONCLUSTERED INDEX [IDX_PakageID] ON [Parts].[Nop_PackageAttribute]
(
    [PackageID] 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 [Customer]
GO
CREATE NONCLUSTERED INDEX [IX_Nop_PackageAttribute_Key] ON [Parts].[Nop_PackageAttribute]
(
    [Key] ASC
)
INCLUDE (   [PackageID],
    [Value]) 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 [Customer]


CREATE TABLE [Parts].[Nop_PackageAttribute](
    [PackageAttributeID] [int] IDENTITY(1,1) NOT FOR REPLICATION NOT NULL,
    [PackageID] [int] NOT NULL,
    [Key] [int] NOT NULL,
    [Value] [nvarchar](max) NOT NULL,
    [CreatedDate] [datetime] NULL,
    [CreatedBy] [int] NULL,
    [ModifiedDate] [datetime] NULL,
    [ModifiedBy] [int] NULL,
    [DeletedDate] [datetime] NULL,
    [DeletedBy] [int] NULL,
 CONSTRAINT [PK_Nop_PackageAttribute] PRIMARY KEY CLUSTERED 
(
    [PackageAttributeID] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [Customer]
) ON [Customer] TEXTIMAGE_ON [PRIMARY]

Nop_AcceptedValuesOption 表

ALTER TABLE [dbo].[Nop_AcceptedValuesOption] ADD  CONSTRAINT [PK_Nop_AcceptedValuesOption] PRIMARY KEY CLUSTERED 
(
    [AcceptedValuesOptionID] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, ONLINE = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [Customer]
GO

CREATE NONCLUSTERED COLUMNSTORE INDEX [_dta_index_Nop_AcceptedValuesOption_5_1669580986__col__] ON [dbo].[Nop_AcceptedValuesOption]
(
    [AcceptedValuesOptionID],
    [AcceptedValuesID],
    [Name],
    [DisplayOrder],
    [Description],
    [CreatedDate],
    [CreatedBy],
    [ModifiedDate],
    [ModifiedBy],
    [DeletedDate],
    [DeletedBy],
    [Is_Split],
    [AcceptedValuesOption_Value],
    [AcceptedValuesOption_Unit]
)WITH (DROP_EXISTING = OFF) ON [Customer]

CREATE NONCLUSTERED INDEX [_dta_index_Nop_AcceptedValuesOption_8_1074154922__K1_3] ON [dbo].[Nop_AcceptedValuesOption]
(
    [AcceptedValuesOptionID] ASC
)
INCLUDE (   [Name]) 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 [Customer]
GO
CREATE NONCLUSTERED INDEX [_dta_index_Nop_AcceptedValuesOption_8_1074154922__K2_K4_1_3] ON [dbo].[Nop_AcceptedValuesOption]
(
    [AcceptedValuesID] ASC,
    [DisplayOrder] ASC
)
INCLUDE (   [AcceptedValuesOptionID],
    [Name]) 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 [Customer]
GO
CREATE NONCLUSTERED INDEX [NonClusteredIndex-20160824-070515] ON [dbo].[Nop_AcceptedValuesOption]
(
    [AcceptedValuesID] 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 [Customer]
GO
CREATE NONCLUSTERED INDEX [NonClusteredIndex-20160829-015901] ON [dbo].[Nop_AcceptedValuesOption]
(
    [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 [Customer]
GO

CREATE TABLE [dbo].[Nop_AcceptedValuesOption](
    [AcceptedValuesOptionID] [int] IDENTITY(1,1) NOT FOR REPLICATION NOT NULL,
    [AcceptedValuesID] [int] NOT NULL,
    [Name] [nvarchar](500) NOT NULL,
    [DisplayOrder] [int] NOT NULL,
    [Description] [varchar](250) NULL,
    [CreatedDate] [datetime] NULL,
    [CreatedBy] [int] NULL,
    [ModifiedDate] [datetime] NULL,
    [ModifiedBy] [int] NULL,
    [DeletedDate] [datetime] NULL,
    [DeletedBy] [int] NULL,
    [Is_Split] [int] NULL,
    [AcceptedValuesOption_Value] [float] NULL,
    [AcceptedValuesOption_Unit] [nvarchar](20) NULL,
    [IsDeleted] [bit] NULL,
 CONSTRAINT [PK_Nop_AcceptedValuesOption] PRIMARY KEY CLUSTERED 
(
    [AcceptedValuesOptionID] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [Customer]
) ON [Customer]

GO

SET ANSI_PADDING ON
GO

ALTER TABLE [dbo].[Nop_AcceptedValuesOption] ADD  CONSTRAINT [DF_IsDeleted]  DEFAULT ((0)) FOR [IsDeleted]
GO

ALTER TABLE [dbo].[Nop_AcceptedValuesOption] CHECK CONSTRAINT [FK_Nop_AcceptedValuesOption_Nop_AcceptedValues]
GO

那么，如何增强update语句更快呢？

Answer 1

添加到 Brendan\xe2\x80\x99s 答案，交换此转换

\n

inner join  Nop_AcceptedValuesOption AVO with(nolock) \nON convert(varchar(20),AVO.AcceptedValuesOptionID)=PA.Value \n

\n

到

\n

inner join  Nop_AcceptedValuesOption AVO \nON AVO.AcceptedValuesOptionID = try_cast(PA.Value as int)\n

\n

您应该能够用 162 个索引查找替换非聚集列存储扫描。如果您没有嵌套循环计划，请尝试inner loop join。

\n

Answer 2

您应该避免在子句中将函数包装在列周围WHERE。通过环绕 a CONVERT，AcceptedValuesOptionIDSQL Server 必须在表中的每一行上运行该函数，然后才能知道该值是否等于PA.Value。

话虽如此，这很可能是WHERE导致您速度缓慢的部分。

convert(varchar(20),AVO.AcceptedValuesOptionID)=PA.Value

我看到有两件事正在发生。

1. 该表中的数据是通过非聚集索引扫描检索的。
2. 然后执行哈希匹配，然后将其与其他两个表的结果组合起来。

查看是否有另一种方式连接到该表，其中连接两侧的列具有相同的数据类型。

另一种选择可能是更改 上的数据类型AVO.AcceptedValuesOptionID。但是，更改现有表中的数据类型是有风险的。您需要确保在此过程中不会破坏其他查询。

此外，如果您知道始终可以PA.value包含 INT 数据的特定场景，则可以考虑创建临时表并将这些行子集插入到临时表中。您可以使用该列作为 INT 数据类型创建此临时表，然后在原始联接中使用该临时表，而不是基表。