Mini X+ 10GbE SMB Performance w/ Scale

[rickyrickyatx]

Has anyone else got an X+ and is attempting to utilize SMB with it? I seem to be maxing out at about 2.5Gb/s reading or writing to two mirrored SSDs in the unit.

Compared to my previous device, this is blazing fast, but a far cry from the 10Gb the network interface is capable of.

If I run the Black Magic Speed test on two separate computers agains the SMB mount at the same time, they both max our around 2.5Gb/s, which makes me think this is a limitation of the CPU in the device and how Samba works (one thread per connection).

Thanks for any advice!

 

[morganL]

Blackmagic is oriented to testing a single local disk and not a network attached array..... you might test with Fio and get better test results.

 

[rickyrickyatx]

I resolved the issue by installing Core 12.x instead of Scale onto my X+. It seems like Scale + SMB has a few odd issues, including SMB crashes, slower I/O (than core), and seemingly lower throughput for a single SMB thread (once again, than core). I also noticed that the web front end was almost non-responsive when I was on Scale, and had an SMB instance pegged at 100% utilization.

Scale is not quite ready for production yet for SMB shares on the X+ IMO, and since I don't need any of the docker / VM / K8 features that come with Scale, moving back to Core did not cause me any issues at all.

There might also be an issue with encrypted volumes and the processor in an X+ on ZFS for Linux, so moving back to Core there provided a nice speed boost to my TimeMachine backups as well, which I store on an encrypted volume.

On a final note, BlackMagic Disk test absolutely is a valid test for a NAS. To whomever runs into this post in the future, if you have odd SMB issues with Scale try downgrading / sidegrading to core and see if the problems persist. That's easier at the moment than attempting to tune or troubleshoot Scale.

 

[morganL]

I resolved the issue by installing Core 12.x instead of Scale onto my X+. It seems like Scale + SMB has a few odd issues, including SMB crashes, slower I/O (than core), and seemingly lower throughput for a single SMB thread (once again, than core). I also noticed that the web front end was almost non-responsive when I was on Scale, and had an SMB instance pegged at 100% utilization.

Scale is not quite ready for production yet for SMB shares on the X+ IMO, and since I don't need any of the docker / VM / K8 features that come with Scale, moving back to Core did not cause me any issues at all.

There might also be an issue with encrypted volumes and the processor in an X+ on ZFS for Linux, so moving back to Core there provided a nice speed boost to my TimeMachine backups as well, which I store on an encrypted volume.

On a final note, BlackMagic Disk test absolutely is a valid test for a NAS. To whomever runs into this post in the future, if you have odd SMB issues with Scale try downgrading / sidegrading to core and see if the problems persist. That's easier at the moment than attempting to tune or troubleshoot Scale.

There has been an identified issue with Linux ZFS encryption on Atom core processors like the Mini. FreeBSD is faster right now.

SMB and Samba also behave a little differently on Linux. Linux seems to handle more clients well, but is not faster on a single client. For the record can you describe the speed differences on your set-up.

I agree with your assessment.. if you don't have need for Linux, K8s or KVM, its best to stay on CORE. SCALE is still maturing and 12.0 CORE is very mature.