How much?
Until I know that, I have absolutely no way to judge whether they are suitable.
Given that I can find a 1.6Tb 9200 for about £800, I'm guessing that new / bigger ones are also stupendously expensive too.
Micron has replaced its U.2 and AIC-format 9200 SSDs with an U.2-only 9300 line, beefing up capacities from 11TB to up to 15.36TB. The 9200 included ECO, PRO and MAX variants for capacity-optimised, read-intensive and mixed workloads, but the ECO model has been scrapped for the 9300. Micron_9300 The 2.5-inch (U.2) 9300 SSD …
If you could live with the appallingly stone-aged SATA, the 5200 ECO has been at ~£700 a couple of weeks ago for the 3.84TB model (although it's risen again recently).
As you say, pricing is everything and I'm worried that the demise of the ECO line will mean there's no longer a dirt cheap 1 DWPD drive for use as fast bulk storage - the PRO line tended to be substantially more expensive than the ECO line. To continue the SATA example, the 5200 ECO 3.8TB is currently selling for about £780, the 5200 PRO 3.8TB for £1100.
Actually, it is capacitor-backed DRAM acting as a write-buffer, hence the wide margin between writes and reads. SLC would not work well, or be anywhere close to as fast, in that role since. These classes of SSDs can have as much as 16GB of DRAM for metadata (LBA maps, etc.,) and write-buffering/coalescing.
For the sake or argument, the 9200 MAX 7.7TB was $3100 online. I doubt the new 16TB model will be over $5000, heck, pretty sure it will be around $3100.
NB: that the 1.6TB 9200 MAX was around $900, which is actually pretty good, considering reliability-performance, imagine that in a RAID 5 or RAID 10 ... can a controller cope with that IO, scrap that, can any interface support 12 of these maxed out IO-wise?
Now everybody repeat after me "RAID is not backup".
On a more serious note it obvioulsy depends on if you want to cater for a (single) device failure or something a bit more catastrophic. If the data is that important then both RAID and (potentially offsite) backup are what you're after. Though all that comes at a price, but how much do you value your data?
RAID is a perfectly adequate backup. As a SINGLE unit. Two drives in a RAID is *NOT* a backup of one of those drive. But one RAID set backing up to another RAID set that you then take offline is a not-unreasonable (and pretty damn fast) backup solution as a single, solitary backup.
It makes for a very timely restore, quick access to single files, and doesn't require huge expense with tape libraries and the like. Most especially, the equipment required to recover it is positively minimal even in the worst case scenario (whereas a new tape library on short notice of a very particular model / compatibility is going to cost you big).
If you have half a brain, you have a "live" RAID set, a "secondary" RAID set to provide redundancy elsewhere, maybe a handful of little offline RAID sets, and if you want then you can have your tapes and other stuff for more "emergency" type restores.
But two drives in a RAID is not a backup. Because each drive only contains (for example) 1 sixth of the full RAID6 set. And in restoration / resyncing, you can easily (in fact it's likely that you'll) lose drives and that means nothing might restore at all.
But having a "copy" RAID set as a quick-restore, poke-around-the-backups, staging and non-critical (No single backup type, unit, media or technology should be "critical" and the only way to restore something! That's the point of a backup!) backup... another RAID set is fine.
Technically, one of my many backups is literally a RAID set on a portable device, which offers the RAID over iSCSI, such that I could - in theory - run every VM and data backup I have off a single portable device without having to do anything more than turn it on and add the iSCSI address to a server. No "18-hour restores". Not saying I rely solely on that for business continuity, but it's saved having to do a full or even partial restore from other technologies and provided direct access to the file from a backup at basically network-speed, many a time.
Geographically dispersed replicated storage simply means that "rm -rf /" deletes files in more locations than you were expecting.
"Local Catastrophes" encompasses more than just disks, servers or server rooms going titsup and telling the finance department they can't have their files back from the Whizbang kamakuza hightech multibuilding distributed storage system because Stubbins their tea boy deleted them (and their snapshots) when he meant to start the robotic kettle on is going to cost both of you your jobs.
"It makes for a very timely restore, quick access to single files, and doesn't require huge expense with tape libraries and the like."
This is an important factor to consider.
Apart from the library costs (robots are cheap), a LTO8 drive to slot into one runs to around $22k, which does offset the $130/tape somewhat.
On the flipside, I trust my tape backups far more than anything done to any other kind of media, and: "a new tape library on short notice of a very particular model / compatibility is going to cost you big" is simply FUD - in an emergency all you need is access to the right kind of tape drive (pulled out of the back of the library and removed from its sled if necessary) - any software which can't handle that is unfit for purpose. Sure, you have to change the tapes manually but a robot is for convenience when doing restores, not an essential piece of kit.