Hi Damien,
First of all, thank you for your time responding to this email. I do not
know what info was supposed to provide, so I am completing it as info is
requested. In this case, I am using a kernel 6.1.0, coming down from
debian stable. The ZFS version that I am using is 2.1.11, and
unfortunately I cannot test it against the raw drive because I would be
destroying my pools (as also Patrick points out).
I have a feeling that the drives are not the ones to blame because a)
they get reported
(https://ssd.userbenchmark.com/SpeedTest/1471/SanDisk-SDSSDP064G) as
performing at at 190+ MBps, way better than I am obtaining, and b) I
bought 2 of these drives on one hand, and the third separately one year
later (so the changes of getting a bad batch are pretty low. If I run
the tests on the pool that has the two drives, in strip mode, I get
double the performance... so the SSD behave consistently in that aspect.
Before coming here I passed by the openzfs libera chat, where some
colleagues tried to help pointing out that could be related to the SATA
controller. With respect to that, I have observed that the other drive
(HDD ST3500412AS) connected to that same SATA controller offers the
following results:
BS bandwidth (MBps) IOPS
4K 93 23700
8K 101 12300
16k 96 5871
32k 97 2970
64k 103 1568
128k 100 788
256k 99 392
which are consistent with the results observed in
https://hdd.userbenchmark.com/SpeedTest/5332/ST3500412AS.
So... everything points to either an evil combination of the SATA
controller and the SSD, or a misconfiguration somewhere in my linux that
I can not pinpoint. The reason because of I landed here is... to try to
gain some knowledge, because professionally I am quite far from storage
system benchmarking, so... I can get numbers, but interpreting those is
definitely another discussion :-/
Thank you for any further insights, if any!
Regards,
---
Felix Rubio
"Don't believe what you're told. Double check."
