PromQL - Detect under-utilized servers

Hello,

I’m trying to use Prometheus data, in order to find all under-utilized servers.
Now, I’m defining a server as under-utilized if :

  • CPU load is below 10% during the last 60 days (on a daily basis average)

OR

  • usage is below 50% and variance is low, during the last 60 days (on a daily basis average)

This is where I am right now, trying to fetch the data for a particular server :

avg_over_time(node_load15{instance="ServerOne",job="events_prometheus"}[60d]) / (number of cores)

But I don’t think this is the right direction, I’m struggling a little bit, I hope you could help me with this one.

Thank you.

Any suggestions regarding this requirements?