I recently read through an interesting White Paper on evaluating the Greenness of a data center put out by the Uptime Institute, written by John R. Stanley, Kenneth G. Brill, and Dr. Jonathan Koomey. You can get it here.
The report outlines four metrics which are designed to help us quantify energy consumption and they present some interesting formulas which I present below:
SI-POM (Site Infrastructure Power Overhead Multiplier) which tells us how much power is consumed to run the DC facilities vs. to run the gear inside it.
Data Center Power consumption at the meter / total hardware consumption at the plug
They suggest that a Tier 4 Data Center will peg out at 2.2
H-POM (Hardware Power Overhead Multiplier) tells us how much power is wasted on power supply conversion, loss to fans, etc. rather than power going to run the components of the gear.
AC hardware load at the plug / DC Hardware compute Load
They suggest that a data center with a 2.2 SI-POM will peg out at 1.33 on this exercise.
Deployed Hardware Utilization Ratio (DH-UR) is the metric that helps quantify the deployed equipment that is drawing power but is comatose; menaing not running an application, but is left on.
Number of servers running live applications / Total number of deployed servers
Number of Terabytes of storage holding data / Total terabytes actually deployed
Deployed Hardware Utilization Efficiency (DH-UE) helps us quantify the opportunity for servers and storage to increase utilization by virtualizing. This conceptually gives us another way of looking at why we would want to consolidate servers at 25% load to run at 50% load in a virtual and scalable framework.
Minimum number of servers necessary to handle peak compute load / Total number of servers deployed
I will spend more time, as I get it, to play around with some numbers to get some real world computations. Feel free to leave your own in the comments section too. The tide raises all boats in our sea of knowledge.