Question

我在存储过程中有一个查询，它从我的SQL Server数据库的几个表中检索数据。从我的应用程序中，我使用一些参数调用此存储过程来过滤数据。

奇怪的是，对于＆＃34; X＆＃34;返回的所有数据一年中的一周（大约2600条记录）检索数据大约需要30秒。但是，如果我添加一个特定的＆＃34;中心＆＃34;过滤一年中的同一周（约1800条记录），获取数据大约需要3分钟！

如果我只运行查询，它运行良好（所有数据为30秒，过滤数据为22秒）。问题是当我通过存储过程运行查询时！

这怎么可能？为什么过滤数据比获取所有数据需要大约x6倍？我为什么失踪？我正确地编写存储过程吗？我怎样才能以更有效的方式做到这一点？

我的存储过程代码是这样的：

ALTER PROCEDURE EventMonitoring
    @EventType AS CHAR(1),
    @Year AS INT,
    @Week AS INT,
    @CenterID AS CHAR(2),
    @AreaID AS INT
AS
    DECLARE @SQLCommand AS VARCHAR(MAX)
    DECLARE @MessageError AS VARCHAR(MAX)

    SET @SQLCommand = ' SELECT ....
                   ....
                   ....
                   ....
                   FROM   Event E
                   INNER JOIN ...
                   INNER JOIN ...
                   INNER JOIN ... 
                   WHERE E.EventYear = ' + CAST(@Year AS VARCHAR) +  
                   ' And E.EventWeek = ' + CAST(@Week AS VARCHAR) +
                   ' And E.EventType = ' + CAST(@EventType AS VARCHAR) 

    IF @Centro <> '-' --If application sends - as the parameter, it gets all centers
    BEGIN
        @SQLCommand = @SQLCommand + ' AND E.CenterID = ''' + @CenterID + '''' 
    END

    IF @Area <> 0 --If application sends 0 as the parameter, it gets all  areas
    BEGIN
        @SQLCommand = @SQLCommand + ' AND E.AreaID = ' + CAST(@AreaID AS VARCHAR) 
    END

    SET @SQLCommand = @SQLCommand + 'GROUP BY ....'

    BEGIN TRY
        EXEC(@SQLCommand)
    END TRY
    BEGIN CATCH
        ....
    END CATCH

Answer 1

无论如何震撼，我敢打赌，可以通过向索引添加额外的列（CenterId）来解决这个问题。

当差异归结为'当我向where子句添加一个额外的列时，查询需要更长的时间'，这表示查询可以使用的索引不包含附加列子句。

要解决此问题，请查找查询使用的现有索引而不使用additional子句，并将附加列添加到该索引的include()，或者创建一个新索引（似乎浪费了一个新的索引来添加但是还有一个int列。

使用快速查询查找Event使用的索引，并查看它是否包含CenterId。如果您不知道如何找到它，您基本上需要检查执行计划。如果您仍然无法找到它，您可以分享您的执行计划，我们将帮助您找到它。使用Paste The Plan @ brentozar.com分享您的执行计划以下是说明：How to Use Paste the Plan。

我只能（模糊地）猜测索引会是什么样子而不了解Event如何加入查询的其余部分，但索引可能看起来像这样，只缺少{{1 }}：

CenterId

Answer 2

我在工作中面临这样的情况，我们只是在proc中添加SET ARITHABORT ON参数，因为ADO连接。

修改

当一个过程在运行时生成脚本时，sql引擎不会看到真正的参数来为查询构建一个好的计划。它调用参数嗅探。所以尝试这样的事情：

ALTER PROCEDURE EventMonitoring @EventType AS CHAR(1), @Year AS INT, @Week AS INT, @CenterID AS CHAR(2), @AreaID AS INT AS DECLARE @MessageError AS VARCHAR(MAX) BEGIN TRY IF @Centro <> '-' --If application sends - as the parameter, it gets all centers BEGIN SELECT .... .... .... .... FROM Event E INNER JOIN ... INNER JOIN ... INNER JOIN ... WHERE E.EventYear = @Year And E.EventWeek = @Week And E.EventType =EventType AND E.CenterID = @CenterID GROUP BY .... END IF @Area <> 0 --If application sends 0 as the parameter, it gets all areas BEGIN SELECT .... .... .... .... FROM Event E INNER JOIN ... INNER JOIN ... INNER JOIN ... WHERE E.EventYear = @Year And E.EventWeek = @Week And E.EventType =EventType AND E.AreaID = @AreaID GROUP BY .... END END TRY BEGIN CATCH .... END CATCH

另一次尝试

在SqlZim指出的文章中快速阅读后，尝试使用ARITHABORT ON和option(recompile)
进行最后一次拍摄
ALTER PROCEDURE EventMonitoring @EventType AS CHAR(1), @Year AS INT, @Week AS INT, @CenterID AS CHAR(2), @AreaID AS INT AS SET ARITHABORT ON DECLARE @SQLCommand AS VARCHAR(MAX) DECLARE @MessageError AS VARCHAR(MAX) SET @SQLCommand = ' SELECT .... .... .... .... FROM Event E INNER JOIN ... INNER JOIN ... INNER JOIN ... WHERE E.EventYear = ' + CAST(@Year AS VARCHAR) + ' And E.EventWeek = ' + CAST(@Week AS VARCHAR) + ' And E.EventType = ' + CAST(@EventType AS VARCHAR) IF @Centro <> '-' --If application sends - as the parameter, it gets all centers BEGIN @SQLCommand = @SQLCommand + ' AND E.CenterID = ''' + @CenterID + '''' END IF @Area <> 0 --If application sends 0 as the parameter, it gets all areas BEGIN @SQLCommand = @SQLCommand + ' AND E.AreaID = ' + CAST(@AreaID AS VARCHAR) END SET @SQLCommand = @SQLCommand + 'GROUP BY .... OPTION (RECOMPILE)' BEGIN TRY EXEC(@SQLCommand) END TRY BEGIN CATCH .... END CATCH

对SQL Server存储过程的查询对于特定数据比对所有数据要长

2 个答案: