Control parallelism in file manager operations

I need to move all the contents of my cache pool to my storage array in order to update the firmware on my NVMe drives (as a precaution more than full necessity), as I'll be removing the drives from the pool and passing them into a Windows VM.

I found the mover was extremely slow, so I opted to use the Unraid file manager to move the contents of /mnt/cache to /mnt/disk3 directly. I also checked "use sparse option" for copying VM qcow2 disk images to preserve the sparseness metadata. While this seemed to be at least slightly faster than the mover, it's still incredibly slow, averaging in the single digit MB/s. Sequential writes to my HDDs are usually over 100 MB/s in comparison.

When I run htop I can see there are at least 16 unique rsync processes running, which is likely a big reason for the slow write speed, with so many concurrent writes to the same disk. The number 16 may be a result of my 16-core CPU. But these file operations are definitely I/O-bound and not receiving any benefit from increased parallelism, rather the concurrent file operations are only slowing each other down.

It would be helpful if there was a way to control this parallelism in file manager operations in some way to improve performance. The default choice also seems naive at best. I'd expect it to take into account the type of disk being written to and read from and avoid concurrent writes to and reads from the same hard disk.

Edited June 6, 20251 yr by Ancalagon