Mario 'BitKoenig' Holbe [off-list ref] wrote:
I have a raid0 over 4x dm-crypt over 4x raid1 (degraded) over 4 disks
...
Since request size is a limiting factor for dm-crypt performance, it
would be great if at least raid1 could also tune itself to the
underlying queue max_hw_sectors (or at least max_sectors).
Just as a small motivation: eliminating raid1 in the above setup gains
me +50MB/s raw dd-throughput (260MB/s vs. 310MB/s). When a filesystem
(xfs) enters the game, the gain can increase even more - I remember
about 300MB/s streaming read on a pure device-mapper based setup
comparable to mine (same hardware) while I'm now stuck at about 210MB/s.
regards
Mario
--
The opposite of a correct statement is a false statement.
But the opposite of a profound truth may well be another profound truth.
-- Niels Bohr