File Server Dfs Replication
DFS is a good solution, and if it's a moderate amount of files it would work well. However, in my experience the initial synchronization can take a really long time if you have many hundreds of thousands of files. You might consider pre-seeding with robocopy. There are several articles about that process, including this one from TechNet: https://technet.microsoft.com/en-us/library/dn495044(v=ws.11).aspx. I've used this process successfully. The robocopy command I wound up with was:
To force an actual replication, use the same command with the SYNCNOW parameter, plus the server to replicate with, the replication group name and a time for which to ignore the schedule (If you've defined a value for it). A DFS namespace is a hierarchy of UNC paths that don’t use fileserver names but transparently redirect users to the nearest copy of the data. Part of this technology is DFS-R, or Distributed File System Replication, which can be used in conjunction with DFS namespaces or as a standalone file replication solution.
File Server Dfs Replication Download
/MIR and /XO means that it will mirror the source to the destination, and will only copy what's changed since you ran it last (handy because you can run it multiple times until it's almost completely caught up, then when you enable DFS there's very little change processing for it to do).
/COPYALL and /DCOPY flags make sure you get the right permissions.
/XD and /XF make sure that you don't pre-seed the folders DFS uses for itself.
/L means log only, always a good idea to run it once with that first, so you know you're not about to screw something up.
Of course, if you're only going to migrate once and not keep DFS going in the future, then this becomes your answer to #4 :)