Can someone please stupid check me on this crazy idea I have?
Problem I am trying to solve:
I have approximately fleventy bajillion SAS and SATA drives kicking around my lab that are perfectly usable, but I just don't want to keep the massive, 6-year-old Dual Xeon servers they are all currently in running 2/47 just to store cold data.
Project goal:
Take this Turing Pi I am supposed to finish ordering and somehow turn it into a low-power NAS that can connect as many drives as possible. The goal is emphatically NOT to be able to run all drives at full speed. The goal is to be able to stand up half a petabyte behind a 1Gbit network connection so that I can warehouse all the copies of peoples' computers and smartphones I need to make whenever someone I know gets a new device, and I inevitably have to both migrate all their data AND apparently keep a backup of that data until the end of time. (
SIGH)
Crazy Idea basic outline:
1) Put a Turing RK1 in Slot 1 of the Turing Pi 2.5. This should give that RK1 module access to a Mini-PCIe slot. This RK1 would boot off of it's dedicated M.2 drive to support the RK1's OS.
2) Put a Mini-PCIe-to-SATA card in the Mini-PCIe slot. Something like this:
https://www.amazon.ca/dp/B0C3C9BWDW/. This should give that RK1 module 4 SATA ports capable of operating in port multiplier/selector mode.
3) Use a SATA hub to expand each of the SATA slots on the Mini-PCIe card from 1 to 5. Something like this:
https://www.amazon.ca/Chenyang-Splitter-Converter-Multiplier-Selector/dp/B0BYN219FL/.
Would this not give me the ability to run 20 SATA drives off of a single RK1 module?
Even Crazier Idea++ outline (now with added cables!):
Module 1: 16GiB RK1, NAS node and Plex client
Has Mini-PCIe slot, so attach 20 SATA drive storage tomfoolery
Has HDMI port, so install a GUI Linux distro
USB 2.0 port to run input devices
Module 2: 8GiB RK1, dedicated NAS node
Has Mini-PCIe slot, so attach 20 SATA drive storage tomfoolery
Module 3: 8GiB RK1, dedicated transcoder node
Has 2x standard SATA ports, to be configured as transcoding cache SSDs.
Module 4: 8GiB RK1, dedicated NAS node
Has 4x USB 3.0 ports, which can each reasonably run 5 SATA drives via
https://www.amazon.ca/dp/B09TG3186B
Thus, if I am not mistaken, I could reasonably run 60 SATA drives, a dedicated transcoder node, and Plex client off of a single Turing Pi 2.5 cluster, using 4x RK1 modules as the compute.
On paper it looks like this should work. Can someone tell me the terrible reason(s) why I won't?
Thank you all in advance!