ArneTR
commented
2 years ago The tool in its current form on Github does orchestrate the software in Docker containers and pulls their metrics via a stream from docker stats.
This approach is just chosen for the prototype. As you pointed out correctly it does not measure energy directly and a problem with this approach is:
- The resolution of 1s is pretty low
- Docker stats gives sometime out different value than top / /proc/stat depending on your OS configuration
- It does not directly translate into energy without an appropriate transfer model.
Scaphandre I know, but since it uses RAPL it has problems of its own like the limitation to Intel CPUs and also a known strong bias and very off dynamic to the real input power signal (https://www.mdpi.com/1996-1073/12/11/2204)
So at the moment this is just our outward facing prototype and we are currently in the process of doing the needed research on our Test-Systems where we setup an electrical measurement directly on the ATX rails of a Blauer Engel for Software compatible machine to create a software transfer model that can either directly translate from sensors / performance metrics to energy values or allow other folks to measure on our system directly (Similar to the Marcher System from Greensoft, only with a docker setup instead of a function base approach (https://greensoft.cs.txstate.edu/index.php/category/documentation/marcher/))
I updated the readme to make the docker step and the current problems more obivious. Thanks for taking the time to have a look at the tool.
I also wrote you an email with some further questions.
hi there.
I'm trying to understand how this metrics tool works, but I'm struggling to understand it from looking at just the code - would you please point me to the documentation for understanding what's happening under the hood?
I've read this: https://www.green-coding.org/blog/green-metrics-measurement-tool/
And I've followed the link to the project demo: https://github.com/green-coding-berlin/green-metric-demo-software
But it's still not obvious to me - are you measuring things like the increase in CPU usage reported by the docker containers?
I've contributed some docs to Scaphandre, a similar project for measuring the energy usage of compute, and you can see how they work out the energy you an attribute to a given process below:
https://hubblo-org.github.io/scaphandre-documentation/explanations/how-scaph-computes-per-process-power-consumption.html
I'm happy to add some more documentation to this project once I understand it 👍