The story has a beginning in 2013. Me and my friend got node (aka physical server) in data center for free by marketing company for a year.
Time was going, a year has ended, but DC owners didn’t do anything with it. So, we lived on node until a couple weeks ago they are remembered about and were asked us to renew the contract with a charge or they’ll shut the server down.
We took a few days to think. Due to we didn’t use full capacity of the node, we stopped the contract and bought the Odroid XU4 instead. Now the blog is working pretty well on this tiny server.
The one thing that annoyed me is a cooler. Periodically, each minute it started for 15s in avg and working with a noise. Quite appreciable “zzzzzzzz”.
The good news is Armbian (it’s used as system on Ordoid) allows to manage it’s cpu activity without reboot. So, in a few hours, I wrote a tool to monitor system loading and managing of cpu. That tool turn some of them off when loading is low (<10.0 for all cpus) in a short time and turn on when loading is high (>60.0) in a few seconds.
The tool gave less temperature on cpu 57℃ vs 66-68℃ (with periodically turning on the cooler) with saved performance.
The code is available on github.