NCQ TRIM Gets Some Improvements With Linux 4.2

[GreatEmerald]

That's just not true. It's better for the SSD longevity if you use TRIM, otherwise the garbage collection routines will increase write amplification as it will maintain the unused stale data.

Hm, since it's been a while, I reviewed the information on this again. Looks like the recommended way of dealing with TRIM is to run fstrim once every week.

From what I can tell, TRIM should be fine to run after every delete, now that it's queued, but it depends on whether the controller behaves properly. If the controller notes which space is no longer used, waits for the percentage to reach fairly high, and then does the consolidation work, then it's perfectly fine. But if the controller does the consolidation work after every single TRIM command, it will be needlessly slow and wear out the drive instead. And there's no real way to tell how the controller behaves. Since you don't need TRIM until the blocks get fairly full, doing it every week is quite a safe bet (unless you constantly write and delete lots of data on it). It also looks like running fstrim after a Btrfs balance is a particularly good idea.

[Ardje]

That's just not true. It's better for the SSD longevity if you use TRIM, otherwise the garbage collection routines will increase write amplification as it will maintain the unused stale data.

Also in the early days. Knowing what does not contain data will enhance the life of your SSD as it does not have to copy "non-data" around. The only reason not to do it is because you have old firmware that doesn't know how to handle it, or you have an SSD that needs all I/O stalled to perform the trim. Anyway: I currently have no performance issues using discard on all my ssd backed filesystems.