So let's see... using a secondhand, probably consumer grade SSD that came with your used server you got off eBay... storing the database on that one SSD, with absolutely no redundancy... I mean, it's probably better than Onion Farms, but that's not saying much.
The database should have been stored on redundant storage. Ideally, multiple SSDs that are rated for the kind of abuse thousands of users using your site would inflict, with a solid RAID solution to ensure no data loss and ideally no restoring from backups.
With ZFS, he could have stuck four SSDs in there and done a four way stripe-mirror with a very minimal write performance hit and up to a 2x read performance boost, with block level checksums, guarranteed file integrity, and automatic healing. Any one SSD could have failed and the site stays up. To recover, you turn off the server, swap the bad SSD, and turn the server back on. The site remains up while ZFS clones the relevant data to the new SSD and then you're done. If you only need the capacity of one SSD, then a four way mirror would give you the same results except that any three SSDs could fail and the site stays up, at the cost of increased wear and tear on the SSDs.
But then again, this is a guy who is so inept that he doesn't even know what server he bought or what it's motherboard has for ports and slots. What he is saying doesn't even make sense anyway, as the board only has one NVMe slot on it. Yes, I'm aware you can get riser cards that adapt one PCIe x16 slot to four M.2 slots, but that would give him five total.