Power usage effectiveness (PUE) is a way to measure how efficiently a data center uses energy. It compares the energy used by computing equipment to the energy used for cooling and other support systems.

PUE is calculated by dividing the total energy used by a data center by the energy delivered directly to computing equipment. A PUE of 1.0 is the most efficient, meaning all energy is used for computing. Any energy used for non-computing tasks, such as lighting or cooling, is considered part of the facility's energy use.

PUE was first created by a group called The Green Grid. In 2016, it became an international standard under ISO/IEC 30134-2:2016.

Issues and problems with the power usage effectiveness

The PUE metric is the most commonly used way to measure how efficiently a data center uses energy. Even though it works better than other methods, it has some problems. Many people in the fields of data center operations, building maintenance, and architecture use PUE to check how well their buildings save energy. Sometimes, workers may forget to include energy used for lighting in their calculations, which can make PUE numbers look better than they are. This mistake is usually caused by human error, not a problem with the PUE system itself.

One major issue is that PUE does not consider the outside temperature of the city where a data center is located. For example, a data center in Alaska, where it is colder outside, may not need as much cooling as one in Miami, where it is hotter. Cooling systems use about 30% of a data center’s energy, while the equipment inside uses nearly 50%. Because of this, a data center in Miami might have a PUE of 1.8, and one in Alaska might have a PUE of 1.7. However, the Miami center might actually be more efficient overall because it needs more cooling.

According to a study in Science Direct, PUE numbers are not useful unless the data center’s computers are working at full power.

To improve PUE calculations, it is important to fix problems like not accounting for outside temperatures. Solving these issues helps ensure that energy use is measured accurately, which can lead to better improvements in data center efficiency.

To get accurate results, all energy data from a data center must be included in PUE calculations. Small mistakes can greatly affect the final PUE number. One common issue is adding energy from alternative sources, like wind or solar power, to PUE calculations. This can make it harder to see the true performance of a data center. Another problem happens when some devices that use energy are shared with other systems outside the data center, which can cause large errors in PUE results.

Benefits and limitation

PUE was introduced in 2006. It was supported by The Green Grid, a group of IT professionals, in 2007. Now, it is the most common way to measure how efficiently data centers use energy. Although it is called "power usage effectiveness," it actually measures the energy use of the data center.

The PUE metric has several benefits:
1. It can be measured again and again over time, allowing a company to track changes in efficiency or during events like seasonal changes.
2. It helps companies see how better practices, such as turning off unused hardware, lower energy use.
3. It creates competition, as companies aim to lower their PUE values. This can be used as a marketing tool to show improvements in efficiency.

However, there are some problems with the PUE metric. The main issue is how the ratio is calculated. Since IT load is the only part of the calculation in the denominator, any decrease in IT load (such as through virtualization or more efficient hardware) can cause the PUE to increase even if the data center is more efficient.

Other problems include the efficiency of the power supply network and accurately measuring IT load. According to Gemma’s analysis, total energy use includes both the energy used by equipment and losses from the power network. PUE is calculated by adding these two and dividing by IT load. This means that power network losses can increase PUE, even if the data center is more efficient. Accurate IT load measurements are also important. Using the actual IT load, not the equipment’s rated power, is crucial. Server usage affects IT energy use and PUE. For example, a data center with a high PUE value but high server usage might be more efficient than one with a low PUE value and low server usage. Some in the industry are concerned that PUE is used as a marketing tool, leading to the term "PUE Abuse."

Power Compute Effectiveness (PCE) has been proposed as a complementary measure. While PUE compares total facility energy to IT equipment energy, PCE compares the available computing power to the electrical capacity of the site. Experts are studying PCE as part of new ways to measure efficiency, especially in environments where computing output is a better indicator of system efficiency than energy use.

Notably efficient companies

In October 2008, Google's data centers had a power usage effectiveness (PUE) ratio of 1.21 across all six locations. This was considered the best possible result at that time. Microsoft followed with a PUE of 1.22.

Since 2015, Switch, the company that builds SUPERNAP data centers, has had a PUE of 1.18 for its SUPERNAP 7 facility in Las Vegas, Nevada. This facility has an average cold aisle temperature of 20.6 °C (69.1 °F) and average humidity of 40.3%. This result is due to Switch’s patented hot aisle containment and heating, ventilation, and air conditioning (HVAC) technologies.

As of the end of Q2 2015, Facebook’s Prineville data center had a PUE of 1.078, and its Forest City data center had a PUE of 1.082.

In October 2015, Allied Control claimed a PUE of 1.02 by using two-phase immersion cooling with 3M Novec 7100 fluid.

In January 2016, the Green IT Cube in Darmstadt was completed with a PUE of 1.07. It uses cold water cooling through the rack doors.

In February 2017, Supermicro announced the use of its disaggregated MicroBlade systems. A Fortune 100 company placed over 30,000 Supermicro MicroBlade servers in its Silicon Valley data center, which had a PUE of 1.06.

Through innovations in liquid cooling systems, the French hosting company OVH achieved a PUE of 1.09 in its data centers in Europe and North America. In 2023, OVH reported an overall PUE of 1.29 for the year.

In 2021, Google reported a PUE of 1.1 for all its global data centers, with some locations achieving a PUE below 1.06.

In 2022, the Research Institutes of Sweden achieved a PUE of 1.0148, which was recorded in the northern part of Sweden.

Standards

PUE was introduced in 2016 as an international standard under ISO/IEC 30134-2:2016 and as a European standard under EN 50600-4-2:2016.