If I'm reading this correctly, there is no redundancy... so if a drive shits the bed, you've lost data... I know having a mix of drive sizes is a pain, but if you have at least multiples of 2 of each size, having a bunch of mirrors and then stripe across would be better. You do lose space, but you gain redundancy. In theory, say 6 drives, 3 different sizes, using zfs. 1 pool full of three mirrors. 500gb, 750gb and 1tb pairs would give you at least some redundancy. But this is just me thinking out loud...
I would look into using ZFS w/redundancy and snapshots esp for backup purposes. This can easily tie into third-party backup services as well unless this is just for transient storage.