My $0.02c worth - I have run all sorts of servers at home over the years, and one of the main challenges around the hardware is managing heat.
I’ve used mini-ITX mobos and tiny cases for builds. They look gorgeous, but at some point, when you stick enough drives in there (assuming you can) or make the CPU/GPU busy, you are going to have a heat problem, or a noise problem, or both.
On my mythtv build I used M-itx and a gorgeous Lian Li small case. It was a beautiful add to my home theatre stack, but in the end I drilled a ton of small holes in the top and added a slow 140mm fan to control the heat without noise.
The same goes for my file server - it was a slightly larger case with no GPU, but once I added my 6th HDD and had a ton of services running, heat became an issue and I was having to add extra fans, which could only be 80mm so they ran fast and noisy.
My new build I’m going to go all the way with a Phanteks Enthoo Full Tower and a few 120mm fans. I’ve decided that looks don’t matter
The other problem for me with these tiny builds is cable management. I’m complete shit at it, and small builds requires some skills. A big case gives you space to spread those cables out.
Lastly, you can get ATX or EATX mobos with 6, 8 or more SATA connectors - room for growth! And there are very low power options available.
I’ll soon have the appleTV + TV upstairs, laptop in the office, and the monster server downstairs with cat-6 + Gb fibre throughout.
Be wary of RAID 5 or 6.
They both have a « write hole » problem (or though much less so in RAID 6). Any power failure which causes an incomplete write can cause a complete RAID corruption - meaning all data is lost. Hardware RAID controllers usually have an onboard backup battery so they can store some information to complete operations should there be a sudden power failure. Software RAID does not have this, and you need to provide a UPS with automatic clean shutdown as the battery runs low using nut or some equivalent.
Some people go as far as to say that RAID 5 should never be used.
You also have very long recovery times when you replace a failed drive (days). Any other failure during this time means total data loss (of course RAID 6 gives you a second redundancy). Weekly Resyncs are very slow too (hours to days), and (unless you constrain your max throughput) will bring your system to its knees.
zfs does not suffer from these problems, BTW.
I run software RAID 5 via mdadm and have a UPS. I’ve replaced drives twice with no issues other than a slightly nervous long wait during recovery. I’m too cheap to buy the extra HDD for RAID 6, and may end up regretting it one day.