Question

有没有办法加速这个SQL？我正在使用链接服务器。我尝试合并，性能相同。

- 通过recordModified Server 1更新参与者 - ＆gt;服务器2

update p1 
set  p1.[RecordCreated] = p2.RecordCreated
,p1.[RecordModified] = p2.RecordModified
,p1.[Email] = p2.Email
from ERSS_Local.dbo.Participant p1
join [Computer2\SqlExpress].ERSS_Local.dbo.Participant p2 
on p2.RecordModified >= (select LastSync from ERSS_DB_Sync) 
and p1.ParticipantID = p2.ParticipantID 
and p1.RecordCreated = p2.RecordCreated and p1.RecordModified < p2.RecordModified

我正在更新超过35列，我将它们留下来简化查询。

我添加了这个并且它似乎没有帮助

 on p2.RecordModified >= (select LastSync from ERSS_DB_Sync)

瓶颈似乎是我正在更新的列数。

是否有加快此更新声明的速度？

Answer 1

本地Sql Server对远程/链接的Sql Server上的统计信息，记录数或索引一无所知。因此，它需要从远程服务器带来所有记录甚至对RecordModified列进行检查。

您可以通过使用`OPENQUERY（）来改进事物，让链接的远程服务器在将结果发送到本地服务器之前预先过滤结果。这将限制网络流量只是相关记录，并允许远程服务器在构建结果集时使用它的索引：

update p1 
    set  p1.[RecordCreated] = p2.RecordCreated
        ,p1.[RecordModified] = p2.RecordModified
        ,p1.[Email] = p2.Email
from ERSS_Local.dbo.Participant p1
join OPENQUERY([Computer2\SqlExpress], 
    'SELECT * FROM ERSS_Local.dbo.Participant  WHERE p2.RecordModified >= ''' + (select LastSync from ERSS_DB_Sync) + '''') p2 
  on p1.ParticipantID = p2.ParticipantID 
   and p1.RecordCreated = p2.RecordCreated 
   and p1.RecordModified < p2.RecordModified

如果您还可以通过将*更改为您需要的内容来限制返回的列，那就更好了。

Answer 2

--assign a variable prevents some issues with query optimizer
DECLARE @LastSync DATETIME;
SELECT @LastSync = LastSync from ERSS_DB_Sync

update 
    p1 
set 
     p1.[RecordCreated] = p2.RecordCreated
    ,p1.[RecordModified] = p2.RecordModified
    ,p1.[Email] = p2.Email
from 
    ERSS_Local.dbo.Participant p1
    join (
            SELECT 
                [Computer2\SqlExpress].ERSS_Local.dbo.Participant  
            WHERE 
                RecordModified >= @LastSync
                -- force the destination engine to filter this set first
        ) as p2
    on p1.ParticipantID = p2.ParticipantID 
    and p1.RecordCreated = p2.RecordCreated and p1.RecordModified < p2.RecordModified

反正有加速这个链接的服务器SQL更新？

2 个答案: