I have a new TS-832PXU with 8 disks in RAID 6 config and when I write a file using "dd" command on Linux on that volume I can see that the raw performance is about 500-600MB/s. I can also write 5 random in parallel (each does 110MB/s then) and when I sum up it's again good performance!
The network interface runs on 2.5GBE (and ethtool confirms speed is 2500) and I see that this limits the effective theoretical performance over network to 250-300MB/s.
However when I access the share via Windows X the transfer speed is per file about 30MB/s only.
If I create multiple parallel threads and count them up 5 threads can exceed 100MB/s and the CPU load goes up to around 60%... but 100Mb/s is still nowhere near 250MB/s.
I lowered the CIFS protocol from smb3 to smb2 and that doubles performance almost to 50MB/s effective file transfer.
Okay, I'm hearing packet signing makes it slower and there's no way to disable in smb3 protocol as it demands it.
Well, I'm unhappy as 30MB/s is really too slow (10% of the NIC and 5% of the raw performance).
When I copy a file from the same client to some other share I have the full performance of over 200MB/s or 7 times faster.
How can I get the best possible performance? Is that QNAPs CPU too weak? (btw, I can't use iSCSI as it's more than 1 client that needs to access)
Need help, please dump your brain
David