Has anyone tested the performance of larger HDDs (14TB+) with a media cache for synchronous writes vs a similar pool with a SLOG device?
I was watching this talk regarding a Ceph retrospective and thought that the implications could apply to ZFS as well.
Here
The TL;DR is that they changed the write behavior from writeback to writethrough and saw a massive reduction in I/O latency due to these rust drives having a write cache built in.