Querying data by joining two tables in two database on different servers

Question

You'll need to use sp_addlinkedserver to create a server link. See the reference documentation for usage. Once the server link is established, you'll construct the query as normal, just prefixing the database name with the other server. I.E:

-- FROM DB1
SELECT *
FROM [MyDatabaseOnDB1].[dbo].[MyTable] tab1
    INNER JOIN [DB2].[MyDatabaseOnDB2].[dbo].[MyOtherTable] tab2
        ON tab1.ID = tab2.ID

Once the link is established, you can also use OPENQUERY to execute a SQL statement on the remote server and transfer only the data back to you. This can be a bit faster, and it will let the remote server optimize your query. If you cache the data in a temporary (or in-memory) table on DB1 in the example above, then you'll be able to query it just like joining a standard table. For example:

-- Fetch data from the other database server
SELECT *
INTO #myTempTable
FROM OPENQUERY([DB2], 'SELECT * FROM [MyDatabaseOnDB2].[dbo].[MyOtherTable]')

-- Now I can join my temp table to see the data
SELECT * FROM [MyDatabaseOnDB1].[dbo].[MyTable] tab1
    INNER JOIN #myTempTable tab2 ON tab1.ID = tab2.ID

Check out the documentation for OPENQUERY to see some more examples. The example above is pretty contrived. I would definitely use the first method in this specific example, but the second option using OPENQUERY can save some time and performance if you use the query to filter out some data.

Try this:

SELECT tab2.column_name  
FROM  [DB1.mdf].[dbo].[table_name_1] tab1 INNER JOIN [DB2.mdf].[dbo].[table_name_2]  tab2   
    ON tab1.col_name = tab2.col_name

If a linked server is not allowed by your dba, you can use OPENROWSET. Books Online will provide the syntax you need.

From a practical enterprise perspective, the best practice is to make a mirrored copy of the database table in your database, and then just have a task/proc update it with delta's every hour.

A join of two tables is best done by a DBMS, so it should be done that way. You could mirror the smaller table or subset of it on one of the databases and then join them. One might get tempted of doing this on an ETL server like informatica but I guess its not advisable if the tables are huge.

If the database link option is not available, another route you could take is to link the tables via ODBC to something such as MS Access or Crystal reports and do the join there.