| 摘要 |
In one embodiment, an upper power limit and an average power limit are specified for each server of a computer system. Power to each server is controlled so that the instantaneous power consumption does not exceed the upper power limit and the average power consumption does not exceed the average power limit. Servers whose average power consumption is currently less than the average power limit are identified. The instantaneous power consumption of each identified server is temporarily allowed to exceed its average power limit, to maximize server throughput. In cases where the average power limit of a device is reduced, such as may occur during peak energy pricing hours, the instantaneous power consumption of the device may be throttled down below the average power limit until the average power consumption no longer exceeds the average power limit.
|
