The 'idle' value refers to the total host idle power, which is not distributed among all the running VMs on the node. Consequently, each VM may wrongly double-count the host's idle time if there are multiple VMs on the same host.
In public cloud environments, it is impractical to determine the number of running VMs on a node. This parameter is intended for use only one bare metal servers that rely on node power estimation because they do not provide real-time power metrics, or when there is only one VM running on the host for very specific scenarios.
So, lets enable this setting through Kepler spec (new feature).
The 'idle' value refers to the total host idle power, which is not distributed among all the running VMs on the node. Consequently, each VM may wrongly double-count the host's idle time if there are multiple VMs on the same host.
In public cloud environments, it is impractical to determine the number of running VMs on a node. This parameter is intended for use only one bare metal servers that rely on node power estimation because they do not provide real-time power metrics, or when there is only one VM running on the host for very specific scenarios.
So, lets enable this setting through Kepler spec (new feature).
Ref: https://github.com/sustainable-computing-io/kepler/issues/973#issuecomment-1760680200