Jan 18 2011

Do the Math with PUE

Calculate the Power Usage Effectiveness (PUE) metric

The Green Grid, an industry trade consortium focused on improving energy efficiency, developed the PUE metric. The metric compares the total amount of energy consumed by the data center with the total amount of energy consumed by the data center’s IT equipment. A rating below 2 is highly effective; above 3, and there are bigger problems than the IT gear.

What It Is: The calculation for PUE is:

PUE = Total data center facility power consumption / IT equipment power consumption

Data Center infrastructure Efficiency (DCiE) is the reciprocal of PUE (1 / PUE), defined as:

DCiE = IT equipment power consumption / Total data center facility power consumption x 100%

Note that The Green Grid defines PUE as being greater than or equal to 1.0, and DciE as being greater than or equal to 100 percent.

What Counts: IT equipment power consumption includes energy used for servers, storage systems, network equipment, monitors, workstations, printers, copiers and other associated technologies. Total data center facility power consumption is measured as the energy going into a facility that supports power distribution units (PDUs), uninterruptible power supplies (UPSes), generators, standby battery power, cooling, lighting and IT equipment.

For example, if a data center facility consumed 3,000 watts (3 kW) of energy, and IT equipment consumed 2,000 watts (2 kW), the PUE would be 1.5 and DCiE 66.667 percent. A PUE of 1.5 (considered to be good) means that energy demands are 1.5 times that of the IT equipment deployed in a given data center. On the other hand, a data center consuming 5 kW of energy with IT equipment consuming 1.5 kW would have a PUE of 3.33 or DCiE of 30 percent, which is considered less effective.

Mileage May Vary: The lower the PUE, the better in terms of how efficiently a facility utilizes energy. Keep in mind that this isn’t an indicator of how effectively energy is used. It’s tempting to focus on the macro metric of PUE and attempt to reduce its value; however, this can be dangerous. The danger lies in trying to lower the PUE without looking at how performance, availability, capacity and economics (PACE) of information services being delivered would be affected. Another pitfall of using PUE for comparisons is an apples-to-oranges scenario in which different time intervals are used for facilities doing various types of work.

Collect Measurements: Total data center input power can be collected via the main meter or metrics supplied by your service provider. IT equipment can be measured from the UPSes, PDUs or meters attached to servers, storage and network gear. Don’t rely on stickers or labels on IT equipment for power consumption information because these may reflect peak or surge versus continuous operating information or may provide fuse and breaker-sizing information.

Maintain Perspective: Metrics are important, but they should be kept in perspective. Multiple metrics are needed — not only the PUE, but also indicators of work being performed or information services delivered. Avoid using dissimilar metrics that don’t reflect your environment or applications workload. Also keep the time interval of the metrics in consideration.

Overall, IT administrators should use the PUE metric along with other measurements to effectively manage their environments. After all, the most important metrics are the ones that provide situational awareness into how your environment is functioning to deliver information services at a given level.

Greg Schulz, author of several books about storage, is founder of the Server and StorageIO Group IT consultancy. His blog can be found at storageioblog.com and at twitter.com/storageio.

aaa 1