logoalt Hacker News

SSDs, power loss protection and fsync latency

64 pointsby ingvelast Wednesday at 9:00 PM7 commentsview on HN

Comments

mgerdtsyesterday at 1:29 AM

It seems pointless to issue flush commands when writing to an NVMe drive with a direct IO implementation that functions properly. The NVMe spec says:

> 6.8 Flush command

> …

> If a volatile write cache is not present or not enabled, then Flush commands shall complete successfully and have no effect.

And:

> 5.21.1.6 Volatile Write Cache

> …

> Note: If the controller is able to guarantee that data present in a write cache is written to non-volatile media on loss of power, then that write cache is considered non-volatile and this feature does not apply to that write cache.

show 3 replies
daymanstepyesterday at 9:19 AM

Did you check that the drives actually honor the flush? Half of drives tested lose FLUSH'd data on power loss.

show 1 reply
jauntywundrkindlast Wednesday at 11:13 PM

Very appropriate topic, after yesterday's High-Performance DBMSs with io_uring: When and How to use it. https://news.ycombinator.com/item?id=46517319 https://arxiv.org/abs/2512.04859

l34zgl45ylast Wednesday at 11:21 PM

[flagged]