

Optional extra, which I do: Change the SSH port to something non-standard. It’s not more secure, but it does cut down on the logspam.


Optional extra, which I do: Change the SSH port to something non-standard. It’s not more secure, but it does cut down on the logspam.


Highly niche requirements not serviced by available hardware without costing waaaay too much. For example, we’re currently using Cisco Meraki MX85, and even they’re not quite there.


Right now I’m using a Fortigate 101E, but eons back I used a simple PC built as cheaply as possible, installed Linux, ip_forward = 1, and iptables.
I am, however, tempted to build something semi-professional based on a banana pi, as I need something custom for work.


Around 700W on low load, almost idle. Haven’t had the chance to test it at full load yet.
It sits in my basement, so the waste heat is actually useful this time of year. Even more so when my basement would otherwise be freezing.
I plan on implementing scheduling where the two most power hungry servers are powered down and up as needed. The other two take almost nothing.
Also, the UPS supports a green mode that is a lot more efficient but with a downside of output side dropping out a few ms during switchover. This removes any conversion loss. I haven’t tested whether my hardware can tolerate it yet, so for now I’m living with the conversion loss.
My house is powered by cheap hydroelectric supply, generated in a mountain roughly 5km from my house, so I’m not worried.
To add to this, I deal with massive storage clusters, and it is my firm professional opinion that hardware raid necessitates a proper raid controller with cache vault. If you have that, then you’re golden. If not, stick to software raid.