System Wide PM policy engine -
This platform independent component would be responsible for implementing
a specified power management policy. The engine would be responsible for
leveraging the various power management mechanisms available on the platform
to honor the objectives specified by the policy. The policy engine could
also interface with other platform independent kernel subsystems to manage
resource power levels and/or provide utilization observability to
higher/lower levels of the PM ecosystem.
Administrative facility -
By default, the system should do something sensible, such as maximizing
performance while minimizing power. In practice, efficiency, throughput,
latency (and even reliability) can trade off...so there may be times when
the default policy is inadequate or suboptimal. For example, one might be
willing to compromise on performance to save power, or vise versa.
This work involves designing and implementing an administrative facility
enabling the expression of PM related objectives.
The facility should provide the administrator with both control and observability.