If your organization/project has data repositories spread across more than one database server, chances are that you will eventually want to pull data from more than one server instance in a single query. If you want to perform these cross-server queries without setting up a persistent Linked Server configuration, a particular configuration option is required.
With regard to your SQL server configuration, this option is known as Ad-Hoc Distributed Queries. In Microsoft SQL Server 2000, the option was enabled by default. This is not the case with Microsoft SQL Server 2005. In order to enable this configuration option, the spgconfigure built-in stored procedure can be utilized like so:
spgconfigure 'Ad-Hoc Distributed Queries', 1 reconfigure
To check that the option has indeed been enabled (and to view your other configuration settings), execute the following statement and verify that the run configuration for Ad-Hoc Distributed Queries shows a value of 1.
Now that you've enabled Ad-Hoc Distributed Queries, you may use either the OpenRowSet or OpenDataSource built-in functions. In the example below, OpenDataSource is used:
select * from OpenDataSource('SQLOLEDB', 'DataSource=servername;User ID=user;Password=password').databasename.schema.tablename
The first parameter passed to OpenDataSource is the OLE driver to be used when making the connection. Here, SQLOLEDB is used for MSSQL Server access; however, you could just as easily pull from an MS Access database (or other data sources for which you have an OLE driver).
The second parameter passed to OpenDataSource is the standard connection string for your database server instance. A fantastic list of these is available from ConnectionStrings.com if you are having trouble building one for your particular data source type.
Once OpenDataSource has been executed, you chain off of it as if you are building a standard query across databases which exist on the same server. If we were querying a table of schema dbo, table name cars from the database automotive on server secondserver with username tester and password testpass, your query would look like this:
select * from OpenDataSource('SQLOLEDB', 'DataSource=secondserver;User ID=tester;Password=testpass').automotive.dbo.cars
...and there you have it: An ad-hoc, cross-server query. Should you wish to use this convention on a more permanent basis, you may consider adding a Linked Server, instead.
One benefit to the Linked Server approach: You can tweak the cross-server connection credentials on a per-login basis. If, for instance, you have three power users who require their own, custom-tailored credentials, you can setup default credentials for everyone else's login that does not appear in your pre-configured list (or map unconfigured logins to the distant server using the local server's current login).