Q. What are the Dfs target-selection methods in Windows Server 2003?

Default target selection: This is the default method, which randomly selects a Dfs target in the requesting computer's local site from the available Dfs targets for the link. If no local targets exist in the requesting site, the target-selection process randomly chooses a target from any site in the forest, regardless of its physical proximity to the requesting computer.

The Windows 2003 site-identification process offers improvements over Windows 2000 Server Dfs site identification. In Win2K Server Dfs, the target-selection process obtains the link-target site by querying the link-target server. However, older OSs such as Windows NT Server 4.0 don't know this information, so the target-selection process in Win2K Server can't identify a site if it includes targets that are NT 4.0 or earlier systems. In Windows 2003, the Dfs server uses the IP address of the target links to determine their location relative to the requesting client, then points the client to a local link target. This method lets the target-selection process recognize older systems (by their IP address) and include them as potential link targets.

Restricted same-site target selection: This option, which also exists in Win2K Server, lets an administrator set Dfs so that clients are never directed to a Dfs target outside of their local site. This restriction solves the problem of clients being directed to targets that are physically far from the client, which would require large amounts of bandwidth, but also means that if the target-selection process can't find a local target for a link, the client can't access the data.

Least-expensive target selection: This is a new method in Windows 2003. You can enable this method as long as the domain controller (DC) that's acting as the Intersite Topology Generator (ISTG) for each site containing Dfs servers is running Windows 2003. When no link targets are available in the local site, this method finds link targets that are "closest" in terms of site costs (i.e., the most efficient path to a target) instead of randomly choosing a target from anywhere in the enterprise. This method is far more bandwidth-efficient than the Default Target Selection method.