Data center PUE: definition and challenges
To meet the growing demand for digitalization, the digital sector uses a large amount of energy, which is a subject of debate. So, to determine whether a data center is efficient, its PUE (Power Usage Effectiveness) must be measured. Here we explain what this indicator calculates, why it is necessary to know the PUE of a data center, and how this indicator can be improved.
What is the data center PUE?
PUE (Power Usage Effectiveness) is an energy efficiency indicator. It was created in 2007 by the Green Grid consortium, which brings together major players in the digital sector. It is used to measure the ratio between the total energy consumed by the entire data center and the energy consumed solely by the IT load, i.e., by equipment such as servers, storage, networks, etc.
Grâce au calcul du PUE d’un data center, il est donc possible de savoir quelle part de l'énergie est perdue dans les infrastructures de support, notamment le refroidissement et les systèmes d'alimentation (comme les onduleurs ou PDU - Power Distribution Units). Ainsi, plus le PUE est élevé, moins la consommation d'énergie d'un data center est efficace. A contrario, plus le PUE se rapproche de 1, plus sa consommation d'énergie est bien gérée.
Calculation and ideal PUE
To calculate this energy efficiency indicator, divide the total energy consumption of the data center by the energy consumption of the IT load. For example, if a data center uses 60,000 kWh of energy, of which 50,000 kWh is used for IT equipment, its PUE would be 1.2. This is an excellent score.
PUE: what are the typical values in data centers?
The ideal PUE for a data center is 1, which means that 100% of the total energy consumed goes to the servers and 0% to support (cooling in particular). But in reality, a very good PUE is estimated to be between 1.2 and 1.4. According to ADEME (1), the average PUE of data centers in France is 1.7, compared to 1.6 on average in the European Union. The goal is to move towards a PUE of 1.2 in new installations.
PUE challenges in a data center
As seen above, the PUE of data centers must be reduced as much as possible. Reducing it is a major challenge on several levels:
● To reduce costs.
A high PUE means high energy consumption. This is one of the biggest expenses in a data center. Reducing PUE therefore improves the energy performance of sites and thus lowers operating costs.
● To combat pollution.
Reducing the energy consumption of facilities automatically reduces greenhouse gas (GHG) emissions. Reducing PUE therefore mitigates the environmental impact of the data center.
● To better manage resources.
Calculating PUE helps highlight energy waste. Site optimization involves improving cooling, which is the largest source of energy consumption after IT equipment.
Understanding relationship between PUE and data center availability
A very low PUE is a sign of quality in terms of energy management, but that's not all. In fact, to achieve excellent energy efficiency (a PUE close to 1.0), data centers use modern, high-performance solutions. They also monitor the temperature of their servers with great precision. It is this quality in the design of the data center, as well as this constant monitoring, that ensures that equipment will not fail due to overheating, for example.
As for the availability of a data center, it relies primarily on the redundancy of electrical and cooling systems. To ensure that servers continue to operate even in the event of a failure or overheating, data centers implement redundant systems. These devices consume energy, which can increase the PUE. In other words, a highly redundant and therefore highly reliable data center can sometimes have a higher PUE than a center optimized solely for energy efficiency. This shows that PUE and availability do not always go hand in hand. A low PUE reflects good energy management, but it does not automatically guarantee high availability.
How can you optimize and reduce a data center's PUE?
Several areas for improvement can be implemented within a data center to reduce Power Usage Effectiveness, for example:
● Improving cooling: this is the most energy-intensive aspect and therefore the first to be rethought. To improve cooling, it is advisable to adopt high-performance techniques such as liquid cooling or free cooling (using cold outside air to lower the inside temperature).
● Virtualization: consolidating several virtual servers on a single physical server reduces the number of active machines and therefore overall energy consumption.
● Use of modern equipment: using newer, more energy-efficient servers and equipment is also a way to improve PUE. Virtualization also allows more functions to be run on fewer physical machines.
● Use of renewable energy: using a renewable energy source to power the data center reduces the site's environmental impact and also meets current energy saving requirements.
● Regular monitoring: PUE must be measured regularly to quickly detect any inefficiencies and implement corrective measures.
PUE is an excellent way to check a site's performance, but other indicators can also be used as a supplement. These include WUE (Water Usage Effectiveness), CUE (Carbon Usage Effectiveness), ERE (Energy Reuse Effectiveness), and REF (Renewable Energy Factor).
Why reduce it?
The advent of artificial intelligence (AI) and the growing need to process ever-increasing amounts of data are placing ever-greater demands on data centers in terms of energy consumption. If PUE remains high, this explosion in energy demand is likely to result in skyrocketing costs and pollution. Reducing PUE therefore guarantees savings on the bill, but also a reduction in environmental impact. In fact, for every kWh avoided, 60 gCO2 are also avoided.
UltraEdge: our commitments to a more sustainable data center
At UltraEdge, we have chosen to focus on energy efficiency by investing heavily in improving our PUE. Where possible, we have also chosen to modernize and upgrade existing infrastructure. This has enabled us to revitalize seven large data centers (in Paris, Lyon, Bordeaux, Strasbourg, Lille, and Rennes).
Our network now consists of more than 250 small-scale edge data centers spread across France. By being anchored in the heart of each region, we are as close as possible to your activities. This translates into better energy consumption management, lower latency, and enhanced security.