White Paper

Dell Power Center’s Power Policies for 12

-Generation Servers

Hence, power over-provisioning is a common issue in power management when using the nameplate

instead of the actual measured power usage values. It is also one of the best and easier ways to first

address power utilization in the data center.

Consider the following example: A server has a nameplate capacity at 500W and a total rack capacity

of 5000W, allowing the administrator to deploy 10 servers. However, when monitored, Power Center’s

power consumption graph reports the average power consumption level of each server to be 150W and

the maximum power consumption level to be 200W for the duration of a week. Therefore 20 - 25

servers could actually be added to the rack. Compared to the 10 servers, the rack density is increased

by 150%. The efficiency rate may be increased even higher if power consumption is monitored and

adjusted over time. In many data centers, expansion is limited because the administrator does not

have the actual information on how much power their operations are using and how much spare power

is available.

Note: The example above is used to illustrate how to address over-provisioning power to a rack. The

numbers listed are used as examples only and the numbers will depend on the types of servers.

Note: For the examples above and below, it is recommended to allow some extra (10-20%) power

capacity due to unforeseen circumstances. If power capping is expected to be immediately effective

when inserting the server in a rack, the server power caps should be set up ahead of time through the

iDRAC or through Power Center before insertion into the rack. It is recommended to power-up servers

in a rack in a staggered fashion to assure no over-subscription to power during bootup.

Note: For the example below, note that only Dell 12

-generation PowerEdge servers currently support

power capping, while Dell 11

- and 12

-generation PowerEdge servers support power monitoring

(PMBus power supplies required).

The following is an example of the power deployment steps administrators would take to control and

optimize the power consumed by servers in a rack:

1. First, set up a Rack group in Power Center and apply a power capping policy of 2000W to that

rack. Once servers are set up in Power Center and become a part of the Rack group, the power

capping policy setup for the rack can help prevent power consumption spikes from tripping the

power breaker.

2. Deploy servers to the rack and start monitoring power consumption. Since the cap is set at

2000W, if the total power usage in the rack is above the cap limit, Power Center will attempt

to maintain the power usage under 2000W and generate power cap alerts in the console.

3. Analyze the power consumption trends of servers in the rack using Power Center.

4. Mitigate and adjust the power policy as needed using Power Center.

Administrators would repeat the last three steps until the optimal power usage is achieved in the rack.

This process can be similarly applied for power deployment to an aisle, room or an entire data center

to achieve an optimal power deployment strategy.

Using Power Policies for Emergency Management

Emergency Power Reduction (EPR) allows the administrator to do the maximum power throttling on a

selected number of servers to minimize their power usage in case of an emergency. For example,

during a critical event, such as a brownout or a cooling failure, the administrator may want to