Since in the cloud it is often not possible to measure energy directly we have created a Machine Learning estimation model based on the data from SPECPower

The setup of the model is based on a research paper from Interact DC and the University of East London.

Our model allows for inline measuring in Watts as well as energy budgeting in Joules with many optional input params to make the model more accurate.

In the chart on the right you can see the performance for an out-of-sample prediciton. Please find more details for in-sample predictions, exploratory data analysis and application documentation on Github.

The model is open-source AGPLv3 Licensed

$ ./static-binary | python3 xgb.py --make intel -- cpu-freq 2600 --ram 7 --cpu-threads 24
191.939294374113
169.99632303510703
191.939294374113
191.939294374113
191.939294374113
191.939294374113
194.37740205685841
191.939294374113
169.99632303510703
191.939294374113
....