I have an external USB hard drive (for backups) on our server. I's shared as "T". Other computer backup to \server\T.
I'd like to also refer to that drive locally as \server\T rather than (say D:) because the drive letter changes when I rotate USB drives.
Is there a performance penalty to referring to it by it's UNC name? (I.e., as if Windows XP were actually going out on the network to get to the drive.)
You will go through the redirector, and I have seen a case when a large file was being referenced in that manner an issue occurred where all network access to the share suddenly failed. In our case we were restoring a SQL Server backup on a fairly large database using the UNC path. This was on Windows 2000. When we changed it to refer to the local drive, the problem went away. There are still some issues with 2003. For instance:
All network share access through the SMB protocol (client-side redirector) may fail on a Windows Server 2003-based computer
An alternative solution is to mount that USB drive under an empty folder on an NTFS volume (using junction points), so the drive can be referred to as C:\Mounted\USB or something similar.
Create the folder, then go into Computer Management > Storage > Disk Management, right-click on the partition and "mount under this folder".
This will not change even if the driveletter assignment does.
Depending on what exactly you are doing you can also run into authentication issues. Using UNC paths in application level services forces the service into impersonation situations which can frequently be show stoppers if the proper trusts are not set up.
For example, something we run into a lot in our SQL Server environment is when accessing the file system via SQL Agent jobs. Using physical drive letter paths works (single hop authentication) but using UNC paths to the same file fails because we don't have trusts set up to allow the impersonation to work.