Most people uses the PUE (Power Usage Effectiveness) as a metric to measure how efficient is their data center. However, increasingly, the DCiE (Datacenter Efficiency) metric is becoming more and more accepted as the metric of measurement.
Both metrics were proposed by the Green Grid in 2007, as a means to estimate the energy efficiency of their datacenters and to compare the results against other datacenters as well as to determine if any energy efficiency improvements need to be made.
The PUE is defined as follows:
PUE = Total Facility Power / IT Equipment Power
The lower the PUE, the more efficient is the facility, while its reciprocal, the DCiE is defined as:
DCiE = 1 / PUE = IT Equipment Power / Total Facility Power x 100%
Do note that DCiE is represented as a percentage. The higher the DCiE percentage, the more efficient is the facility.
So, what is the flaw or misrepresentation of PUE? Lets take an example. Assume you are managing a data center facility which has an estimated PUE of 3.4, and currently housing a few thousands of IT equipment which comprises of servers, storage, network devices, etc. Now, assume that you (and a project team) have assessed, analyzed and determined that majority of the servers are underutilized. This paints a very coherent picture as underutilized assets (especially if there are a lot of them) makes the data center inefficient. So, assumes the following:
Total Facility infrastructure power consumption = 3.0 MW
IT servers power consumption = 0.95 MW
Storage devices power consumption = 0.25 MW
Networking devices and other IT assets power consumption = 0.05 MW
PUE = (3.0 + 1.25) / 1.25 = 3.4
Now, assuming your project team architects a virtualization blueprint using a combination of hardware partitioning and software virtualization technologies like VMware to consolidate the servers. Lets say this blueprint, after implementation, will reduce the overall IT servers footprint by 60% and the IT servers power consumption (post-implementation) will be just 0.5 MW. Now, without making any changes to the facility’s cooling and power distribution, all of a sudden, the utilities bill will be reduced because instead of consuming 0.95 MW for IT servers, you’re now down to 0.5 MW only.
However, if we re-calculate the PUE again, it will be higher. The PUE = (3.0 + 0.8) / 0.8 = 4.75. Therefore, even when your IT footprint have shrunk, all things being equal, the PUE actually went up!
Where else, if you look at it from a DCiE perspective, the original DCiE will be 41.67%, and the post-consolidation DCiE will be 26.67%. Although the percentage drops, it does show the shrinking power consumption for IT equipment as compared to the facility power consumption.
The moral of the story here are two-folds: Firstly, the metric to be used must be clear, precise and not be confusing. You will need to baseline current state prior to implementing changes and progressively measure the changes to track improvements or impact from the changes.
Secondly, IT and facility are never and should never be treated in isolation. Changes to IT will impact and cause repercussions to the facility and its capacity, and likewise changes to the facility will impact the IT assets within it. Both needs to be aligned and as marriage counselors would say, only when both parties works together, will the marriage be a fruitful and happy union.
Tags: Consolidation, Data Center, DCiE, Density, Efficiency, PUE, Virtualization