I just got a Seagate 8TB SMR drive and before using it as backup, I wanted to create huge files to play around with.
I wrote a small program that will create a binary file of a certain size, I used it to create a 1TB file. During write, I would get around 240MB/s write speed on my 6 drive RAIDZ2 array, but when the file was created, my volume didn't grow by the expected 1TB size. Instead, when I looked at the compression ration, it was at about x114. The dataset was set to LZ4 compression.
I tried same procedure, but this time I changed compression on dataset to OFF, and this time I was able to run at write speed of about 500MB/s it seems.
I believe the drives still had headroom to go faster.
On average, especially when I access my data over the network, my drives are not breaking a sweat while still maxing out my 1Gbe network.
In a nutshell, overall performance is bound by the various hardware and software limitation.
For instance ZFS on one array will only make use of 1 thread, compression will use another thread, but both will be RAM and CPU bound and under utilizing disk performance.
It seems, if you only want to access your drive over the network, then 6 disk vs 7 will not bring you significant system performance. It will bring you more capacity or better resiliency.
If you really want to achieve mind blowing performance then you may want or look into stripped vdev, I think, and go multi RAIDZx configuration.
