You can design your FTL to be resilient to arbitrary power loss, as long as the NAND chips don't physically go off corrupting unrelated data on power down. That only requires extremely minimal capacitance. I believe Apple SSDs do have some of that in the NAND PMIC next to the storage chips themselves; it probably knows to detect falling voltage rails and trigger a stop of all writes to avoid any actual corruption due to out-of-spec voltages.
I've absolutely heard storage vendors talking about protecting just the FTL during power loss as PLP. You could have an FTL where any writes are atomic, but that gets in the way of throughput practically. The storage vendors don't seem to generally be on board that tradeoff except for 'industrial' branded SKUs that also make throughput tradeoffs anyway.
That's not necessarily true, it just comes down to the design. Apple seem to use a log-structured FTL which works great for performance and just requires a (very fast) log replay after a hard shutdown. You can see the syslog messages from the NVMe controller (via RTKit) talking about rebuilding the table when this happens.
Micron also talks about power loss resistant FTL design in their whitepaper, and although their older SSDs had caps, I think their recent ones mostly do away entirely.
You can design your FTL to be resilient to arbitrary power loss, as long as the NAND chips don't physically go off corrupting unrelated data on power down. That only requires extremely minimal capacitance. I believe Apple SSDs do have some of that in the NAND PMIC next to the storage chips themselves; it probably knows to detect falling voltage rails and trigger a stop of all writes to avoid any actual corruption due to out-of-spec voltages.
I've absolutely heard storage vendors talking about protecting just the FTL during power loss as PLP. You could have an FTL where any writes are atomic, but that gets in the way of throughput practically. The storage vendors don't seem to generally be on board that tradeoff except for 'industrial' branded SKUs that also make throughput tradeoffs anyway.
That's not necessarily true, it just comes down to the design. Apple seem to use a log-structured FTL which works great for performance and just requires a (very fast) log replay after a hard shutdown. You can see the syslog messages from the NVMe controller (via RTKit) talking about rebuilding the table when this happens.
Micron also talks about power loss resistant FTL design in their whitepaper, and although their older SSDs had caps, I think their recent ones mostly do away entirely.
https://www.micron.com/-/media/client/global/documents/produ...
This kind of stuff has been done in filesystems for ages (journaling etc.); there's no reason why FTLs can't be designed the same way.
1 reply →