SQL Server:查询速度快,但程序缓慢
查询运行速度很快:
DECLARE @SessionGUID uniqueidentifier
SET @SessionGUID = 'BCBA333C-B6A1-4155-9833-C495F22EA908'
SELECT *
FROM Report_Opener
WHERE SessionGUID = @SessionGUID
ORDER BY CurrencyTypeOrder, Rank
子树成本:0.502
但是,在存储过程中放入相同的SQL运行速度很慢,并且执行计划完全不同
CREATE PROCEDURE dbo.ViewOpener @SessionGUID uniqueidentifier AS
SELECT *
FROM Report_Opener
WHERE SessionGUID = @SessionGUID
ORDER BY CurrencyTypeOrder, Rank
EXECUTE ViewOpener @SessionGUID
子树成本:19.2
我跑了
sp_recompile ViewOpener
它仍然运行相同(严重),我也改变了存储过程
CREATE PROCEDURE dbo.ViewOpener @SessionGUID uniqueidentifier AS
SELECT *, 'recompile please'
FROM Report_Opener
WHERE SessionGUID = @SessionGUID
ORDER BY CurrencyTypeOrder, Rank
再回来一次,试图欺骗它重新编译。
我删除并重新创建了存储过程,以便让它生成一个新的计划。
我试过使用一个诱饵变量强制重新编译, 并防止参数嗅探 :
CREATE PROCEDURE dbo.ViewOpener @SessionGUID uniqueidentifier AS
DECLARE @SessionGUIDbitch uniqueidentifier
SET @SessionGUIDbitch = @SessionGUID
SELECT *
FROM Report_Opener
WHERE SessionGUID = @SessionGUIDbitch
ORDER BY CurrencyTypeOrder, Rank
我也试着用WITH RECOMPILE
定义存储过程:
CREATE PROCEDURE dbo.ViewOpener @SessionGUID uniqueidentifier
WITH RECOMPILE
AS
SELECT *
FROM Report_Opener
WHERE SessionGUID = @SessionGUID
ORDER BY CurrencyTypeOrder, Rank
所以它的计划永远不会被缓存,并且我试着在执行时强制重新编译:
EXECUTE ViewOpener @SessionGUID WITH RECOMPILE
哪些没有帮助。
我已经尝试将过程转换为动态SQL:
CREATE PROCEDURE dbo.ViewOpener @SessionGUID uniqueidentifier
WITH RECOMPILE AS
DECLARE @SQLString NVARCHAR(500)
SET @SQLString = N'SELECT *
FROM Report_OpenerTest
WHERE SessionGUID = @SessionGUID
ORDER BY CurrencyTypeOrder, Rank'
EXECUTE sp_executesql @SQLString,
N'@SessionGUID uniqueidentifier',
@SessionGUID
哪些没有帮助。
实体“ Report_Opener
”是一个未被索引的视图。 该视图仅引用基础表。 没有表格包含计算列,索引或其他。
对于它的地狱,我试图创建视图
SET ANSI_NULLS ON
SET QUOTED_IDENTIFER ON
这并没有解决它。
那是怎么回事?
我尝试将视图的定义直接转移到存储过程中(违反了3个业务规则,并且打破了重要的封装),并且仅使其速度降低了大约6倍。
为什么存储过程版本如此之慢? 与其他类型的特别SQL相比,运行临时SQL的SQL Server可能会更快吗?
我真的不想
完全改变代码
Microsoft SQL Server 2000 - 8.00.2050 (Intel X86)
Mar 7 2008 21:29:56
Copyright (c) 1988-2003 Microsoft Corporation
Standard Edition on Windows NT 5.2 (Build 3790: Service Pack 2)
但是,如果SQL Server无法像运行查询的SQL Sever那样快速运行,那么究竟是什么原因,如果不是参数嗅探的话。
我的下一次尝试是将StoredProcedureA
调用StoredProcedureB
调用StoredProcedureC
调用StoredProcedureD
来查询视图。
如果失败了,请让存储过程调用存储过程,调用UDF,调用UDF,调用存储过程,调用UDF来查询视图。
总而言之,以下QA运行速度很快,但放入存储过程时速度较慢:
原本的:
--Runs fine outside of a stored procedure
SELECT *
FROM Report_OpenerTest
WHERE SessionGUID = @SessionGUID
ORDER BY CurrencyTypeOrder, Rank
sp_executesql
:
--Runs fine outside of a stored procedure
DECLARE @SQLString NVARCHAR(500)
SET @SQLString = N'SELECT *
FROM Report_OpenerTest
WHERE SessionGUID = @SessionGUID
ORDER BY CurrencyTypeOrder, Rank'
EXECUTE sp_executesql @SQLString,
N'@SessionGUID uniqueidentifier',
@SessionGUID
EXEC(@sql)
:
--Runs fine outside of a stored procedure
DECLARE @sql NVARCHAR(500)
SET @sql = N'SELECT *
FROM Report_OpenerTest
WHERE SessionGUID = '''+CAST(@SessionGUID AS varchar(50))+'''
ORDER BY CurrencyTypeOrder, Rank'
EXEC(@sql)
执行计划
好计划:
|--Sort(ORDER BY:([Expr1020] ASC, [Currencies].[Rank] ASC))
|--Compute Scalar(DEFINE:([Expr1020]=If ([Currencies].[CurrencyType]='ctCanadianCash') then 1 else If ([Currencies].[CurrencyType]='ctMiscellaneous') then 2 else If ([Currencies].[CurrencyType]='ctTokens') then 3 else If ([Currencies].[CurrencyType]
|--Nested Loops(Left Outer Join, OUTER REFERENCES:([Openers].[OpenerGUID]))
|--Filter(WHERE:((([Currencies].[IsActive]<>0 AND [Currencies].[OnOpener]<>0) AND ((((((([Currencies].[CurrencyType]='ctUSCoin' OR [Currencies].[CurrencyType]='ctMiscellaneousUS') OR [Currencies].[CurrencyType]='ctUSCash') OR [Currencies].
| |--Nested Loops(Left Outer Join, OUTER REFERENCES:([Currencies].[CurrencyGUID], [Openers].[OpenerGUID]) WITH PREFETCH)
| |--Nested Loops(Left Outer Join)
| | |--Bookmark Lookup(BOOKMARK:([Bmk1016]), OBJECT:([GrobManagementSystemLive].[dbo].[Windows]))
| | | |--Nested Loops(Inner Join, OUTER REFERENCES:([Openers].[WindowGUID]))
| | | |--Bookmark Lookup(BOOKMARK:([Bmk1014]), OBJECT:([GrobManagementSystemLive].[dbo].[Openers]))
| | | | |--Index Seek(OBJECT:([GrobManagementSystemLive].[dbo].[Openers].[IX_Openers_SessionGUID]), SEEK:([Openers].[SessionGUID]=[@SessionGUID]) ORDERED FORWARD)
| | | |--Index Seek(OBJECT:([GrobManagementSystemLive].[dbo].[Windows].[IX_Windows]), SEEK:([Windows].[WindowGUID]=[Openers].[WindowGUID]) ORDERED FORWARD)
| | |--Clustered Index Scan(OBJECT:([GrobManagementSystemLive].[dbo].[Currencies].[IX_Currencies_CurrencyType]))
| |--Clustered Index Seek(OBJECT:([GrobManagementSystemLive].[dbo].[OpenerDetails].[IX_OpenerDetails_OpenerGUIDCurrencyGUID]), SEEK:([OpenerDetails].[OpenerGUID]=[Openers].[OpenerGUID] AND [OpenerDetails].[CurrencyGUID]=[Currenc
|--Hash Match(Cache, HASH:([Openers].[OpenerGUID]), RESIDUAL:([Openers].[OpenerGUID]=[Openers].[OpenerGUID]))
|--Stream Aggregate(DEFINE:([Expr1006]=SUM(If (((([Currencies].[CurrencyType]='ctMiscellaneous' OR [Currencies].[CurrencyType]='ctTokens') OR [Currencies].[CurrencyType]='ctChips') OR [Currencies].[CurrencyType]='ctCanadianCoin') OR [
|--Nested Loops(Inner Join, OUTER REFERENCES:([OpenerDetails].[CurrencyGUID]) WITH PREFETCH)
|--Nested Loops(Inner Join)
| |--Index Seek(OBJECT:([GrobManagementSystemLive].[dbo].[Openers].[IX_Openers_OneOpenerPerSession]), SEEK:([Openers].[OpenerGUID]=[Openers].[OpenerGUID]) ORDERED FORWARD)
| |--Clustered Index Seek(OBJECT:([GrobManagementSystemLive].[dbo].[OpenerDetails].[IX_OpenerDetails_OpenerGUIDCurrencyGUID]), SEEK:([OpenerDetails].[OpenerGUID]=[Openers].[OpenerGUID]) ORDERED FORWARD)
|--Index Seek(OBJECT:([GrobManagementSystemLive].[dbo].[Currencies].[PK_Currencies_CurrencyGUID]), SEEK:([Currencies].[CurrencyGUID]=[OpenerDetails].[CurrencyGUID]) ORDERED FORWARD)
坏计划
|--Sort(ORDER BY:([Expr1020] ASC, [Currencies].[Rank] ASC))
|--Compute Scalar(DEFINE:([Expr1020]=If ([Currencies].[CurrencyType]='ctCanadianCash') then 1 else If ([Currencies].[CurrencyType]='ctMiscellaneous') then 2 else If ([Currencies].[CurrencyType]='ctTokens') then 3 else If ([Currencies].[Currency
|--Nested Loops(Left Outer Join, OUTER REFERENCES:([Openers].[OpenerGUID]))
|--Filter(WHERE:((([Currencies].[IsActive]<>0 AND [Currencies].[OnOpener]<>0) AND ((((((([Currencies].[CurrencyType]='ctUSCoin' OR [Currencies].[CurrencyType]='ctMiscellaneousUS') OR [Currencies].[CurrencyType]='ctUSCash') OR [Currenc
| |--Nested Loops(Left Outer Join, OUTER REFERENCES:([Currencies].[CurrencyGUID], [Openers].[OpenerGUID]) WITH PREFETCH)
| |--Filter(WHERE:([Openers].[SessionGUID]=[@SessionGUID]))
| | |--Concatenation
| | |--Nested Loops(Left Outer Join)
| | | |--Table Spool
| | | | |--Hash Match(Inner Join, HASH:([Windows].[WindowGUID])=([Openers].[WindowGUID]), RESIDUAL:([Windows].[WindowGUID]=[Openers].[WindowGUID]))
| | | | |--Clustered Index Scan(OBJECT:([GrobManagementSystemLive].[dbo].[Windows].[IX_Windows_CageGUID]))
| | | | |--Table Scan(OBJECT:([GrobManagementSystemLive].[dbo].[Openers]))
| | | |--Table Spool
| | | |--Clustered Index Scan(OBJECT:([GrobManagementSystemLive].[dbo].[Currencies].[IX_Currencies_CurrencyType]))
| | |--Compute Scalar(DEFINE:([Openers].[OpenerGUID]=NULL, [Openers].[SessionGUID]=NULL, [Windows].[UseChipDenominations]=NULL))
| | |--Nested Loops(Left Anti Semi Join)
| | |--Clustered Index Scan(OBJECT:([GrobManagementSystemLive].[dbo].[Currencies].[IX_Currencies_CurrencyType]))
| | |--Row Count Spool
| | |--Table Spool
| |--Clustered Index Seek(OBJECT:([GrobManagementSystemLive].[dbo].[OpenerDetails].[IX_OpenerDetails_OpenerGUIDCurrencyGUID]), SEEK:([OpenerDetails].[OpenerGUID]=[Openers].[OpenerGUID] AND [OpenerDetails].[CurrencyGUID]=[Cu
|--Hash Match(Cache, HASH:([Openers].[OpenerGUID]), RESIDUAL:([Openers].[OpenerGUID]=[Openers].[OpenerGUID]))
|--Stream Aggregate(DEFINE:([Expr1006]=SUM([partialagg1034]), [Expr1007]=SUM([partialagg1035]), [Expr1008]=SUM([partialagg1036]), [Expr1009]=SUM([partialagg1037]), [Expr1010]=SUM([partialagg1038]), [Expr1011]=SUM([partialagg1039]
|--Nested Loops(Inner Join)
|--Stream Aggregate(DEFINE:([partialagg1034]=SUM(If (((([Currencies].[CurrencyType]='ctMiscellaneous' OR [Currencies].[CurrencyType]='ctTokens') OR [Currencies].[CurrencyType]='ctChips') OR [Currencies].[CurrencyType]='
| |--Nested Loops(Inner Join, OUTER REFERENCES:([OpenerDetails].[CurrencyGUID]) WITH PREFETCH)
| |--Clustered Index Seek(OBJECT:([GrobManagementSystemLive].[dbo].[OpenerDetails].[IX_OpenerDetails_OpenerGUIDCurrencyGUID]), SEEK:([OpenerDetails].[OpenerGUID]=[Openers].[OpenerGUID]) ORDERED FORWARD)
| |--Index Seek(OBJECT:([GrobManagementSystemLive].[dbo].[Currencies].[PK_Currencies_CurrencyGUID]), SEEK:([Currencies].[CurrencyGUID]=[OpenerDetails].[CurrencyGUID]) ORDERED FORWARD)
|--Index Seek(OBJECT:([GrobManagementSystemLive].[dbo].[Openers].[IX_Openers_OneOpenerPerSession]), SEEK:([Openers].[OpenerGUID]=[Openers].[OpenerGUID]) ORDERED FORWARD)
糟糕的是,他们急于卷起600万行; 另一个不是。
注意:这不是关于调整查询的问题。 我有一个快速运行的查询。 我只想让SQL Server从存储过程中快速运行。
我和原来的海报有同样的问题,但引用的答案并没有解决我的问题。 查询仍然从存储过程中运行得非常慢。
我在这里找到了另一个答案“参数嗅探”,谢谢Omnibuzz。 归结为在存储过程查询中使用“局部变量”,但阅读原始文件以获得更多理解,这是一个很好的写法。 例如
慢速方式:
CREATE PROCEDURE GetOrderForCustomers(@CustID varchar(20))
AS
BEGIN
SELECT *
FROM orders
WHERE customerid = @CustID
END
快速方法:
CREATE PROCEDURE GetOrderForCustomersWithoutPS(@CustID varchar(20))
AS
BEGIN
DECLARE @LocCustID varchar(20)
SET @LocCustID = @CustID
SELECT *
FROM orders
WHERE customerid = @LocCustID
END
希望这可以帮助别人,这样做会将我的执行时间从5分钟减少到大约6-7秒。
我发现这个问题,下面是存储过程缓慢版本和快速版本的脚本:
dbo.ViewOpener__RenamedForCruachan__Slow.PRC
SET QUOTED_IDENTIFIER OFF
GO
SET ANSI_NULLS OFF
GO
CREATE PROCEDURE dbo.ViewOpener_RenamedForCruachan_Slow
@SessionGUID uniqueidentifier
AS
SELECT *
FROM Report_Opener_RenamedForCruachan
WHERE SessionGUID = @SessionGUID
ORDER BY CurrencyTypeOrder, Rank
GO
SET QUOTED_IDENTIFIER OFF
GO
SET ANSI_NULLS ON
GO
dbo.ViewOpener__RenamedForCruachan__Fast.PRC
SET QUOTED_IDENTIFIER OFF
GO
SET ANSI_NULLS ON
GO
CREATE PROCEDURE dbo.ViewOpener_RenamedForCruachan_Fast
@SessionGUID uniqueidentifier
AS
SELECT *
FROM Report_Opener_RenamedForCruachan
WHERE SessionGUID = @SessionGUID
ORDER BY CurrencyTypeOrder, Rank
GO
SET QUOTED_IDENTIFIER OFF
GO
SET ANSI_NULLS ON
GO
如果你没有发现差异,我不会责怪你。 根本不在存储过程中。 这种差异将一个快速的0.5成本查询转换成一个包含6百万行的热切假脱机:
慢: SET ANSI_NULLS OFF
快速: SET ANSI_NULLS ON
这个答案也可以做出来,因为视图确实有一个连接子句,它说:
(table.column IS NOT NULL)
所以有一些NULL
参与。
通过返回查询分析器并运行,进一步证明了这一解释
SET ANSI_NULLS OFF
。
DECLARE @SessionGUID uniqueidentifier
SET @SessionGUID = 'BCBA333C-B6A1-4155-9833-C495F22EA908'
。
SELECT *
FROM Report_Opener_RenamedForCruachan
WHERE SessionGUID = @SessionGUID
ORDER BY CurrencyTypeOrder, Rank
查询很慢。
所以问题不在于查询是从存储过程运行的。 问题是企业管理器的连接默认选项是ANSI_NULLS off
,而不是ANSI_NULLS on
,这是QA的默认选项。
Microsoft在KB296769中承认这一事实(错误:不能使用SQL Enterprise Manager创建包含链接服务器对象的存储过程)。 解决方法是在存储过程对话框中包含ANSI_NULLS
选项:
Set ANSI_NULLS ON
Go
Create Proc spXXXX as
....
为您的数据库执行此操作。 我有同样的问题 - 它在一个数据库中工作正常,但是当我使用SSIS导入(而不是通常的恢复)将此数据库复制到另一个时,这个问题发生在我的大多数存储过程中。 所以在google搜索一下之后,我找到了Pinal Dave的博客(顺便说一下,我遇到了他的大部分帖子,并且非常感谢Pinal Dave)。
我在我的数据库上执行下面的查询,它纠正了我的问题:
EXEC sp_MSforeachtable @command1="print '?' DBCC DBREINDEX ('?', ' ', 80)"
GO
EXEC sp_updatestats
GO
希望这可以帮助。 只是传递帮助我的其他人的帮助。
链接地址: http://www.djcxy.com/p/62239.html