
Pros (keep doing):
- Stackable
- Low 54 USD price per HDD slot
- High throughput with 1 Gbps per HDD
- Horizontally scalable storage (with the right software, i.e. seaweedfs, ceph et.al.)
Cons (improve upon):
- 32-bit CPU
- Low bariety of OS support (not that many distros makers release ARM 32-bit Linux distros)
- No OS with atomic updates supports it (i.e. Fedora CoreOS)
3 things missing to make it even better:
- ARM64 CPU (Would open up OS support to Flatcar Linux, AWS BottleRocket, Ubuntu Core and soon Fedora CoreOS)
- OS-es with atomic updates eliminates partially completed upgrades and lowers the consequence of broken upgrades, and eases the burden of patching large amounts of machines.
- Power over Ethernet (PoE) with IEEE IEEE 802.3bt support for single cable attachment.
- With built-in support for PoE we can basically just attach the Ethernet cable to the ODROID-HC3, insert the MicroSD and HDD, and be going.
- WIth a PoE managed switch we also automatically get remote power off, on and cycle, one of the most used features of BMCs on enterprise server boards.
- Lowers the price per HDD if you've already got a 802.3bt capable PoE switch.
- If the board with the HDD attached consumes less than 24W, then IEEE 802.3af PoE support is sufficient.
- Network bootable with PXE boot.
- Can eliminate need for MicroSD card to boot (less cost)
- Will spend less time flashing MicroSD cards for new boards
Adding those features will increase the cost a bit. The future Kobol64 product (https://kobol.io) comes out at a cost of about 65 USD per disk slot (295 EUR / 5 slots, converted to USD), whereas the ODROID-HC2 costs 54 USD per disk slot, so there is still a possibility to beat that price point. For ODROID there's no need to spend time creating a 5-slot cabinet, just develop a new board, keep the HC2 aluminium frame and ship it.

What do you think? Am I barking up the wrong tree? Let me know in the comments below.