If the OS is configured with DVFS the power curve is strongly non linear in relation to load.
This should be a parameter in the data that we train the model with.
After making this availabe in the source data we would also need a flag to feed into the model as possible input parameter.
This parameter is very important because some machines, like the Github VMs, run only with fixed CPU frequency.
If the OS is configured with DVFS the power curve is strongly non linear in relation to load.
This should be a parameter in the data that we train the model with.
After making this availabe in the source data we would also need a flag to feed into the model as possible input parameter.
This parameter is very important because some machines, like the Github VMs, run only with fixed CPU frequency.