Your post is woefully lacking in details. You say you want it to be 'fast', but what makes you think Curl has some performance advantage over other methods?
1) Are the 'numerous' files you need to tranfer on lots of different systems, or just one? If only one, then you will not gain much performance with concurrency.
2) Do you know the sizes of the files at the point when you want to do the tranfer? And the bandwidth to each system? If you have both, you could calculate an optimal N groupings of files. There are numerous CPAN ftp modules. You could either create multiple threads, or spawn multiple processes to handle this N groupings.