For converting large datasets, my usual rule is "Use SQL when you can, use Perl when you have to". Perl is terrific for coding transformations (more fun too), but there's just no substitute for doing work in the server, without moving the data out and back in.

The last couple of times I've had to do things like this, I developed the transforms in SQL, but used Perl to control the execution. It's a good combination that allows you to exploit the strengths of both tools.

Something else to consider: if this is new territory for you, use it as an opportunity to get good with SQL. Try to avoid cursors and row-by-row processing: use set-oriented operations as much as possible. Avoid "procedural SQL".

Another tip: Back up everything before you start! ;-)

HTH