Those scale performance results are pretty shocking, I wonder if that applies to SMB as well. Hopefully IX systems is looking into that. Send like those results are similar to ones that @LTS_Tom did several months ago without much improvements.
I got similar results for scale didn’t run against core 12 though. Wonder what the results would be if containers and VMs were also running? At this point I would not recommend Scale as a data store, maybe as a application platform with locally VM and container store. Think I might experiment a little once I done with current project. Any thoughts on the reasons?
@LTS_Tom Have you ever tested NAS file share performance in TrueNAS Core/Scale? i.e. How fast are most common file operations on SMB/NFS file share?
The Phoronix tests seem to represent like large file copy operations and databases I/O, but not so much a traditional shared file system operations, like searching for files, deleting files, unpacking archives etc that we all do regularly.
My FreeNAS Mini XL (first version) SMB file share performance is very poor so I did a bit of benchmarking on my own. Results were rather surprising. It took over an hour to unpack Linux kernel on SMB file share but less than four minutes on the same NFS share. See my benchmarks at TrueNAS forum for benchmark results performed with a linux kernel archive with six pretty common tasks. I provided sample commands for Windows and Linux/FreeBSD/Unix.
I would be keen to know TrueNAS Scale compares at those tasks vs Core. I would not be surprised if Samba on Linux performs much better for this use case as it is probably more optimised on Linux wiht more kernel offloaded operations. On Core those tasks are CPU bound, so if you run them on the new TrueNAS Mini XL+ I could make a decision if it is time to upgrade…
Understood. It would be an interesting benchmark however, so I hope it moves up on your to-do list . It will be unlike many others, yet, more applicable I to a TrueNAS user who use NAS as file share and not as block share.