White Paper

ManualsBrandsDell ManualsSoftwareOpenManage Power Center Version 1.0

Dell Power Center’s Power Policies for 12

-Generation Servers

servers; however, once excess power is available (known as headroom), the allocation of that power is

distributed considering priority.

Let’s look at an example: Group G1 with 3 servers: S1, S2 and S3, and the group capping is set at

2000W. S1 is set at high priority, while S2 and S3 are set at medium priority. The table below

demonstrates how the power capping is adjusted after each sampling interval (T1, T2, etc.). Power

capping (P1 column) of each server at each interval is calculated based on the power consumption from

the previous interval (P2 column) and is dynamically adjusted as needed. While adjusting the capping

value on each server, Power Center also maintains the total power consumption of the group (row G1,

column P2) within the group capping ceiling (row G1 and column P1). The worst-case scenario is when

the power consumption reaches the budget (interval T3). In this case, an alert is sent, and the

administrator will need to take appropriate mitigating action (for example, to increase the group

capping to 2500W as in T4 column).

The power is dynamically adjusted among servers in a group Table 1.

500W

800W

1000W

1500W

1200W

500W

700W

500W

550W

800W

700W

300W

500W

400W

450W

700W

600W

1300W

2000W

1700W

2000W

3000W

2500W

700W

300W

The table also displays the excess power (row E) or the difference between the power usage and the

power cap values. Excess power will be dynamically distributed among the servers in the group based

on the priority of each server, with the higher-priority server receiving more power. At interval T1,

the excess power of the group is 700W. At the next interval (T2), the power caps are adjusted

accordingly. With higher priority, S1 is allocated 300W from the excess, while S2 and S3 each get

200W. At some point during this interval, the power demand in the group increases from 1300W to

1700W, and S1 power consumption reaches its cap. In the next interval (T3), Power Center will take

the excess power (from S2 and S3) and redistribute this power to S1. In interval T3, each server has

reached its cap limit, and consequently the group has also reached its cap limit. The S1 server will not

be allocated more power (even though it has the higher priority) until the cap is increased or the

workload in the group is reduced to create some power headroom. The capping value is increased from

2000W to 3000W at the T5 interval. The power is now dynamically redistributed among the servers.

The graph below illustrates how the power is dynamically distributed among the servers in the group

and how the total power usage is maintained under the capping value.