I think it comes down to how reliable or fast your internet is. You can selfhost a lot of stuff with reliable internet, power, and environmental variables.
If you have any of those which are less than reliable, a VPS like Linode may make sense for uptime.
My UPS systems have built in monitoring.
Yes, this seems to happen more frequent with brands like realtek. All hardware has failure rates, generally speaking more expensive and enterprise gear fails with less frequency, but can still fail. Personally, I don’t enjoy hardware failure, so I invest in stable clean power and great hardware. It may cost more in the overall power budget but is less headache because things just don’t fail as often.
I sit around 5kw average… Depending what the GPUs are doing
Mine are running all of the time, including during power outages, and are only shut down for physical maintenance and reboot for software maintenance.
This is a little variable through. Windows hosts tend to require more frequent software reboots in my experience. About once a year, I physically open each device and inspect, clean dust (fairly rare to find it for my setup though), and perform upgrades, replace old storage devices and such. Otherwise I leave them alone.
I usually get about 5-7 years out of the servers and 10 out of networking hardware, but sometimes a total failure occurs unexpectedly still and I just deal with it as needed.
No rack mount options? Not for my lab…
Interesting. I’ve been in IT for over 30 years and have always used UPS systems. Never had data loss as a result of power issues. Also, it cleans the power a bit and I get a lot of life from my hardware. Plus, I dont have to worry at all, it works as intended. My entire lab has redundant UPS systems backed by a large generator … Zero downtime is nice.