I put 4 Sandisk 3D 4TB SSDs in the lower bays of the QNAP TS-932PX-4G NAS. My first thought was to configure them as caching disks, but I thought I’d do some measurements with the 4 SSDs in a RAID 10 array (striped and mirrored).
I decided to use a Lightroom catalog directory, which contains live catalogs (lots of small files) and Lr backups (larger files).
I couldn’t get the 10 GbE multimode fiber SFP+ (small form-factor pluggable) module out of the failed QNAS box (NAS1), and I’m waiting on some more, so I performed the testing with a copper 2.5 GbE connection from the Ethernet switch to the NAS, and 10 GbE copper connections from the computer to the switch. In both casse4s the source of the backup was a 32 TB PCIe SSD array.
Backing up to the spinning rust RAID 10 array:
Backing up to the SSD RAID 10 array:
About half again as fast using the SSD array. Not very impressive. I’ll test again with a 10 Gb/s connection, but the speeds I’m seeing here are so far from the wire speed of the 2.5 Gb/s copper connection that I’ve got my doubts.
Here’s the SSD RAID 10 array with 10 Gb/s Ethernet:
About the same speed. It was faster for the large files — about 500 MB/s, but essentially the same speed for the small ones.
If we look at the measured sequential transfer rates, we see that the SSDs are about twice as fast as the magnetic disks.
That translates into smaller, but significant differences for large files: