HP ML150 HP Power Capping and Dynamic Power Capping for ProLiant servers techn - Page 22

Figure 12. Provisioning with Enclosure Dynamic Power Capping Power Provisioning with Enclosure Dynamic Power Capping 7836 Watts (PSU Specification) P ower P rovisioned to P S U • 2612 x 3 = 7836 watts • S ingle 3Ø L ine = 8640 • 16 B lades per 8KW 4000 Watts (Measured Peak) Power Provisioned to Cap at Measured Peak • ≈4KW enclosure • 32 Blades per 8KW 0 Watts Using Dynamic Power Capping and Enclosure Dynamic Power Capping to manage data center power provisioning is examined in greater detail in the ―Dynamic Power Capping TCO and Best Practices‖ white paper available at http://h71028.www7.hp.com/ERC/downloads/4AA2-3107ENW.pdf Additional uses for power capping Both basic Power Capping and Dynamic Power Capping can be used in other capacities in the data center, including emergency management and automatic power control during peak demand periods Power capping for emergency management Power capping can be used to effectively manage server power consumption when unforeseen circumstances arise. A common example would be a cooling system failure in part of a data center. In this situation, administrators can manually lower the power cap on a group of servers using IPM and power capping. This will quickly and efficiently lower server power consumption and heat generation in the affected area until cooling can be restored. Under these circumstances, it may be appropriate to set power caps significantly lower, even though they may impact performance. Using SIM, an administrator can define and store groups of servers as collections. This allows the administrator to quickly apply power caps to these groups in an emergency. Power capping can also be used effectively if all or part of a data center loses primary AC power. An administrator can immediately apply a group power cap to lower power consumption for server groups, decreasing the power drain on data center Uninterruptible Power Supplies (UPSs). This increases the maximum window of time, commonly referred to as ride-through time, the data center can remain operational after a power failure but before power and cooling is restored by the on-site generators. 22