Hi, just a quick question regarding expected cache pool performance in raid0.

I did a quick test in my second unraid tower to test for performance of a raid0 2x ssd's setup vs a single ssd setup. Acording to AI it shoud improve a lot. My actual test results were:

• single cache ssd=> write: 449MB/s read: 541MB/s

root@X:~# dd if=/dev/zero of=/mnt/casiopea/testfile bs=1G count=5 oflag=dsync conv=fdatasync
5+0 records in
5+0 records out
5368709120 bytes (5.4 GB, 5.0 GiB) copied, 24.239 s, 221 MB/s
 
root@X:~# dd if=/mnt/casiopea/testfile of=/dev/null bs=1G count=5 iflag=direct
5+0 records in
5+0 records out
5368709120 bytes (5.4 GB, 5.0 GiB) copied, 9.86418 s, 544 MB/s

So I am getting similar or even worse results.

I am wondering what results I should really obtain. I mean how any pool created via Pool-Devices (not the cache one for /system and /appdata) but a dedicated one only for /domains should perform.

Also, would it improve if I setup the mentioned raid 0 on a dedicated HBA and then access it from Unassigned Devices and not from the array-Pool-Devices?

Hope to find some answers soon.

dd is not a good way to test that; for more real-world results, use, for example, pv:

pv /mnt/cache/path/to/large/file > /dev/null

Make sure you restart the array before each test, so that the file is not read from RAM.

8 hours ago, JorgeB said:

ure you restart the array before each test, so that the file is not read from RAM.

Hi, I did another test and the results are not what I expected for a 2xssd raid 0:

pv /mnt/cache/Superman.2025.REPACK.2160p.UHD.Blu-ray.mkv > /dev/null

60.0GiB 0:03:38 [ 281MiB/s] [=======================================================>] 100%

What is to be expected if i raid0 on an HBA and then mount it via Unassigned Devices? Or what is the maximum I should expect from the fastest setup?

I think looking at it again maybe I have not setup my raid0 as it should. Can anyone confirm if this is ok:

btrfs fi usage /mnt/casiopea/
Overall:
    Device size:                 894.26GiB
    Device allocated:             14.06GiB
    Device unallocated:          880.20GiB
    Device missing:                  0.00B
    Device slack:                    0.00B
    Used:                          7.25GiB
    Free (estimated):            884.96GiB      (min: 444.86GiB)
    Free (statfs, df):           884.96GiB
    Data ratio:                       1.00
    Metadata ratio:                   2.00
    Global reserve:                9.20MiB      (used: 16.00KiB)
    Multiple profiles:                  no

Data,RAID0: Size:12.00GiB, Used:7.24GiB (60.30%)
   /dev/sdg1       6.00GiB
   /dev/sdf1       6.00GiB

Metadata,RAID1: Size:1.00GiB, Used:9.02MiB (0.88%)
   /dev/sdg1       1.00GiB
   /dev/sdf1       1.00GiB

System,RAID1: Size:32.00MiB, Used:16.00KiB (0.05%)
   /dev/sdg1      32.00MiB
   /dev/sdf1      32.00MiB

Unallocated:
   /dev/sdg1     440.10GiB
   /dev/sdf1     440.10GiB

That's slow, you may want to try both SSDs individually to confirm they perform similarly and/or ZFS stripe (raid0)

On 2/5/2026 at 7:35 AM, JorgeB said:

That's slow, you may want to try both SSDs individually to confirm they perform similarly and/or ZFS stripe (raid0)

ZFS raid0: 812MiB/s

Btrfs raid0: 637MiB/s (pv command and 805MiB/s (dd command).

I guess the problem with my previous data was raid0 of 2ssd's was connected to old raid card (running @x4 on old pcie plus only sata ports of 3Gbs).

Please con you confirm this are the results I should obtain?

Rgds and thx for support!

10 hours ago, luca2 said:

ZFS raid0: 812MiB/s

Btrfs raid0: 637MiB/s (pv command and 805MiB/s (dd command).

That looks about right; typically, you would get a little less than the speed of both combined.