What target & actions for this powercapping?

[solonfan]

Hi,
I have a questions about this feature(power capping) design? Whats the target for this feature?The CPU or the total system?
From source,it seems is focus on the system powercapping.If so,what is the action if the power consumption output value out of the power capping rang.Is there any **recommended actions or spec that we can follow?
I am not sure if any more devices that can be power consumption reduced beside CPU.

Thanks,
SolonFan

[cjcain]

The power cap takes into account the total system power. The OCC monitors the total system power.
If the cap is enabled and the system consumption goes above the user defined cap, the OCC will reduce processor frequency/voltage to try to keep the power under that cap. The processors are generally one of the highest power consumption devices in the system, but there are other indirect effects. If the processor power is reduced, the processor temperature should also decrease, which should also lower fan speeds, saving additional power.
The minimum allowed power cap is defined to be the minimum power required to run the system, so normally it should be able to hold that cap unless there are hardware issues causing excessive power consumption.
If the OCC is not able to reduce the power below the user power cap after multiple attempts, an unrecoverable PEL will be logged to inform the operators that the power cap could not be held and the system is moved to safe mode (with reduced power and performance).

From the P10 EnergyScale Whitepaper:

Power Capping

Power Capping enforces a user-specified limit on power consumption. See User Interfaces for
supported interfaces to set a power cap. In most data centers and other installations, when a server is
installed, a certain amount of power is allocated to it. Generally, the amount is what is considered to be
a “safe” value, and it typically has a large margin of reserved, extra power that is never used. This is
called the margined power. The main purpose of the power cap is not to save power but rather to give
a data center operator the capability to reallocate power from existing systems to new systems by
reducing the margin assigned to the existing servers. That is, power capping gives an operator the
capability to add extra servers to a data center which previously had all available power allocated to its
existing systems. It does this by guaranteeing that a system will not use more power than assigned to it
by the operator. This is also called a “hard power cap”.
Previously, the data center administrator had to plan for the power consumption of the data center
based on the Underwriters' Laboratories (UL) rating on the back of the servers being installed. The UL
rating (commonly referred to as “label power”) on today's servers indicates the most power that a
system could ever draw and is based on the capacity of the power supplies. It has to take into account a
fully-configured system with the highest power-usage parts installed at the highest possible utilization