"Enterprise ssd - why not just use a raid of cheap consumer ones?"
RAID-Z3 of them, actually (Or ZRaid1) - You don't need the L2ARC, but they still benefit from the ZIL SLOG drive to smooth out the random writes (with a long enough delay period a lot of random writes end up evaporating entirely instead of turning into sequential ones and the ZIL is only ever used in anger when recovering from an unexpected power cut)
ZFS - the filesystem based on the premise that "Disks are crap, just deal with it" - instead of trying to take the usual enterprise approach of gold plating everything to get extra reliability it never assumes that what you put in is what you get out.
"But enterprise ssds? Seems to mostly be higher write cycles tolerance, and even on consumer level ssds the endurance is "lots" "
This in spades. People are sniffing at 0.5DWPD on 4TB SSDs - forgetting that in the case of sata drives that's getting close to "all writes, all the time" - given that mechanical drives max out at around 120MB/s sequential that's effectively filling them 5-6 times faster than you can do with a mechanical drive, nonstop, over 3-5 years (depending on the warranty period) - this simply isn't a real-world use cases for drives of this scale and means you're unlikely to ever run out of spare blocks _even if_ you decide to use the drives for surveillance systems and don't allocate more spare space just in case.
Biting the hand that feeds IT
