I have added a simple sync write test into current napp-it (Pools > Benchmark > syncwrite).
It opens a file, writes a character and closes the file and checks how often this can be done in an intervall like 60s and gives a good impression of your sync write performance degration.
With a diskbased pool the difference of sync=disabled (use writecache only) vs sync=always (commit every write) is really really huge, even with a good SSD only pool you get a degration.
example (SSD Pool, Sandisk Extreme, 2 x vdev Z2)
It opens a file, writes a character and closes the file and checks how often this can be done in an intervall like 60s and gives a good impression of your sync write performance degration.
With a diskbased pool the difference of sync=disabled (use writecache only) vs sync=always (commit every write) is really really huge, even with a good SSD only pool you get a degration.
example (SSD Pool, Sandisk Extreme, 2 x vdev Z2)
Code:
begin writing..
results
filesystem: /av/test
sync setting: always
write per commit: 1KB
writes: 8691
time (s): 30
writes/s: 289
begin writing... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
results
filesystem: /av/test
sync setting: disabled
write per commit: 1KB
writes: 898385
time (s): 30
write /s: 29946
Last edited: