問題描述

我正在通過鏈接服務器從 VIEWS 中提取大量數據.我使用的是 SQL Server 2012，鏈接服務器是 SQL Server 2008

I am extracting large amount of data via linked server from VIEWS. I am using SQL Server 2012 and linked server is SQL Server 2008

我的選擇語句是

SELECT * INTO MY_LOCAL_TABLE
FROM 
(    SELECT * FROM LINKEDSERVER.DB.TABLE.VIEW
     WHERE DATE>'2012-01-01' AND ID IN (SELECT ID FROM MY_LOCAL_VIEW) 
) Q

我期待 300K 行有將近 700 個 ID.以前需要幾個小時，但現在需要 20 多個小時！！

I am expecting 300K rows for nearly 700+ IDs. before it used to take couple of hours but now its take more than a 20 hr!!

您能否為此 PAIN 提出任何替代解決方案?

Could you please suggest any alternative solution for this PAIN??

非常感謝！

推薦答案

當您使用由 4 部分組成的名稱時，例如 [server].db.dbo.table，尤其是在 中join，很多時候整個表都通過網絡復制到本地機器上，這顯然不太理想.

When you use a 4-part name such as [server].db.dbo.table, especially in a join, often times the entire table is copied over the wire to the local machine, which is obviously not ideal.

更好的方法是使用 OPENQUERY -- 在源(鏈接服務器)處理.

A better approach is to use an OPENQUERY -- which is handled at the source (linked server).

試試:

SELECT *
FROM OPENQUERY([LINKEDSERVER], 'SELECT * FROM DB.TABLE.VIEW WHERE DATE>'2012-01-01')
AND ID IN (SELECT ID FROM MY_LOCAL_VIEW) 

使用這種方法，鏈接服務器將返回日期 > x 的所有行，然后本地服務器將通過本地表中的 ID 過濾該行.

With this approach the linked server will return all rows for date > x, and then the local server will filter that by ID's in your local table.

當然，索引仍然會影響 SELECT * FROM DB.TABLE.VIEW WHERE DATE>'2012-01-01.

Of course, indexing will still play a factor for doing SELECT * FROM DB.TABLE.VIEW WHERE DATE>'2012-01-01.

我在大型子集上使用的另一種方法是將本地 ID 轉儲到遠程服務器，然后遠程處理它，例如:

Another approach, which I use on large subsets, is to dump the local ID's to the remote server, THEN handle it all remotely, such as:

    -- copy local table to linked server by executing remote query 
    DECLARE @SQL NVARCHAR(MAX)
    SET @SQL = 'SELECT ID INTO db.dbo.tmpTable FROM [SERVER].DB.DBO.MY_LOCAL_VIEW'
    EXEC(@SQL) AT [LINKEDSERVER]

   -- index remote table?!?
    DECLARE @SQL NVARCHAR(MAX)
    SET @SQL = 'CREATE INDEX [IXTMP] ON db.dbo.tmpTable (ID)'
    EXEC(@SQL) AT [LINKEDSERVER]

    -- run query on local machine against both remote tables
    SELECT *
    -- INTO sometable
    FROM OPENQUERY([LINKEDSERVER], 'SELECT * 
                                    FROM DB.TABLE.VIEW
                                    WHERE DATE>''2012-01-01''
                                    AND ID IN (SELECT ID FROM db.dbo.tmpTable)')

    -- now drop remote temp table of id's
    DECLARE @SQL NVARCHAR(MAX)
    SET @SQL = 'DROP TABLE db.dbo.tmpTable'
    EXEC(@SQL) AT [LINKEDSERVER]

如果本地視圖也很大，那么您可以考慮執行一個使用 openquery 返回本地機器的遠程查詢(假設遠程機器有本地作為鏈接).

If the local view is also large, then you may consider executing a remote query that uses an openquery back to the local machine (assuming the remote machine has the local as a link).

-- copy local table to linked server by executing remote query 
DECLARE @SQL NVARCHAR(MAX)
SET @SQL = 'SELECT ID INTO db.dbo.tmpTable FROM OPENQUERY([SERVER], ''SELECT ID FROM DB.DBO.MY_LOCAL_VIEW'')'
EXEC(@SQL) AT [LINKEDSERVER]

這篇關于SQL 鏈接服務器查詢非常非常慢的文章就介紹到這了，希望我們推薦的答案對大家有所幫助，也希望大家多多支持html5模板網！