I did some measuring in the past 2 days. With representative workload running on it, one of my 3 home infrastructure Kubernetes nodes uses 13,5 watts on average. πŸ‘πŸ»

Some follow up on this one. Measurement has now going on for 241 hours (~10 days) and the consumption of one node around 13.7 KW/h on average.

The difference can be explained by the shift in workloads throughout the cluster.

I guess in order to provide better data, I'll soon buy a new "energy monitor"(?), this one is (almost) 20 years old by now, that allows me to scrape metrics that I can correlate directly with the workloads on the host.

Recommendations would be great!

Follow

Little correction: it's 13.7 W/h not KW/h.

Β· Β· 1 Β· 0 Β· 0

@sheogorath it is probably 13.7W (not W/h). Watt is "Energy per Time" already.

@claudius in this case it's intended, since it's an average over an hour.

@sheogorath 13.7W can absolutely be the average over any given timeframe or instant.

So I guess you measured somewhere around 650Wh over two days, right? that is on average 13.7W (650 Wh divided by 48h). But even if you would have measured for 15 minutes, the calculation would be the same: 3.4Wh divided by 0.25h would also equal 13,7W.

@sheogorath so the Energy counter rises steadily (so-and-so many watt-hours), but the Power you need is constant (13.7Watt) or at least averages to that amount of power, even if there are spikes.

@sheogorath let's look at it the other way round: you have a 1.000W toaster oven.

If it runs for a second, you just consumed 1.000Ws or about 0,3Wh. Leave it running for one hour, you consumed 1.000Wh. Leave it running for a full day, you consumed 24.000Wh

All while needing 1000W of power in any given moment or timeframe.

Sign in to participate in the conversation
Sheogorath's Microblog

This is my personal microblog. It's filled with my fun, joy and silliness.