The VMguy

What is Distributed Power Management you ask?  DPM is used in your Virtual Infrastructure when the physical hosts are at very low utilization levels.  During the night or on the weekends, if I have 8 nodes in a cluster and each of the nodes are running at 10% utilization, that is not a very good use of resources.  DPM will vMotion (Live Migrate) your virtual machines to a subset of hosts and power down the hosts that are not needed.  I do not have many customers doing this with server workloads.  I do see a lot more interest in DPM from customers using Virtual Desktops.  Think about it, if I run 8 physical hosts to run all of my desktops in my virtual infrastructure, what happens at 5pm on Friday?  At any given point on the weekend how many users are logged into the desktops in my virtual infrastructure?  Maybe 10-20?  Why run 8 physical hosts to handle that workload?  vMotion them automatically to two hosts for redundancy and turn off the rest.  As users login and the workloads increase, Virtual Center will automatically power on additional nodes and spread the workloads across all of the hosts available.  Much like adding cylinders to a car engine while crusing down the highway.  One item to note, this feature is still experimental support from VMware.  On a personal note, I’ve never known an "experimental" feature from VMware to not make it to full support at some point in time.

This new whitepaper has some great info about DPM and how it determines when to shutdown cluster hosts.  Most importantly, and something I have been looking for and found nowhere else, beginning on page 5 it lists all of the advanced options for DPM.  Things like the minimum amount of hardware to leave running and the minimum star rating (much like DRS) that you need before hosts will be shutdown.

Some good reading if your preparing your infrastructure for DPM down the road.  Definately something all of us should use if we can, somebody at our company pays the power bill don’t they?