Architecting a vSphere Compute Platform : Resource Balancing and Transparent Maintenance : 9.3 Distributed Power Management
   
9.3 Distributed Power Management
The vSphere DPM feature significantly reduces the power consumption used in the data center by consolidating virtual machines so that the minimum number of ESXi hosts run during periods of low utilization, such as overnight. DPM works in conjunction with DRS to migrate virtual machines from hosts with low utilization to reduce the number of required hosts, powering down unused hosts automatically, which in turn reduces power consumption through both lower hardware power draw and lowered costs associated with air-conditioning units.
Figure 34. vSphere Distributed Power Management
 
The DPM mechanism takes advantage of remote power-on and power-off technologies such as Intelligent Power Management Interface (IPMI) or out-of-band remote access cards such as the HP Integrated Lights-Out (iLO) or, alternatively, the Wake-on-LAN (WoL) functionality embedded into network interface cards.
Service providers benefit from DPM where workloads vary significantly over time, reducing operational costs with only a minimal initial setup overhead. The design decision to support DPM in the infrastructure requires the service provider to consider:
Purchasing server hardware that includes an iLO or IPMI network interface.
Only running DPM during non-business hours because consumers of service might not want to shut down servers during business hours.
If the host platform does not support iLO or IPMI, purchase NICs for the vSphere vMotion network that supports WoL functionality, although there are configuration prerequisites to use WoL.
Configuring the vSphere DPM automation level for automatic operation, and using the default vSphere DPM power threshold, which decreases power and cooling costs, and administrative overhead.