RAID just shifts the single
babgvantRAID just shifts the single point of failure from the drive to a controller. IMO, it’s not worth the added complexity and risk unless going with a much more expensive controller (i.e. no built in or cheap addin cards). I don’t know how true it is currently, but there was a time when many of the lower end controllers relied on the host CPU for parity calculations and lacked sufficient cache so in some cases it was actually slower than a JBOD config too. FWIW, I JBOD and use the synctoy to fake folder duplication across spindles. Pretty sure Michael (Mikinho) uses RAID on his system, hopefully he can chime in and maybe provide some perspective from the other side 🙂
I do think fewer, larger drives is the way to go. Not only is it a better value, but leaves room and capacity for future expansion when you need it (and you can never have too much storage :)).