07-30-2019, 02:49 PM
(07-30-2019, 12:28 PM)stuartiannaylor Wrote: http://www.iocrest.com/en/product_details440.html
Prob a picex4 to pciex1 packet switch with basically 4x Marvell 88SE9215
With most decent large HD having approx 200MBs still quite a bottleneck and with the overhead of 10 disk RAIDZ2 I doubt you will even get 100MBs per disk.
That is if its recognised and works.
I understand very well that I'd probably have less than 100MB/s per disk with that controller, but as I said it's a RAIDZ2 (stripped data on 8 disks with two redundancy) so it still gives me about 800MB/s data speed on the ZFS datapool (if the RK3399 is powerful enough to deal with that much data computing the redundancy and blocs checksums...).
It will be a NAS, and only a NAS as every other services such as webserver will be deported onto several dedicated Raspberry Pi 4, connected on my gigabit home network. That means I technically only need about 100MB/s data speed on the datapool. So the controller will be more than enough to sustain gigabit data transfers.
The real question is, is that specific SATA controller working properly on a ROCKPro64? Or have someone a better option for connecting 10 to 12 SATA on the ROCKPro64? I'd like to be sure it works before spending several hundreds of dollars on that setup...
For information, my current setup is an old (from 2007) Tyan motherboard with a PCI-X SATA controller supporting all the drives, which has a total bandwidth four times smaller that the PCI-E 2.0 4x... The local access onto the ZFS datapool is about 250MB/s, and it's still enough to saturate my gigabit link. The setup with the ROCKPro64 can only be better than that, even with the bottleneck of the SATA controller.