The escalating energy consumption of data centers is a pressing issue, with significant environmental and cost implications, spotlighting Power Usage Effectiveness (PUE) as a pivotal metric for measuring efficiency. This focus on PUE underscores the progress being made in reducing energy waste and enhancing the sustainability of data center operations through better awareness and practices.

Power Usage Effectiveness (PUE) is a measure of data center energy efficiency, calculated by dividing total facility power by IT equipment power. A PUE near 1 signifies high efficiency, with most power used for computing, while higher PUEs indicate greater power use for cooling and electrical losses.

Dgtl Infra explores Power Usage Effectiveness (PUE), offering an in-depth look at data center efficiency. We provide a thorough examination of how PUE is calculated, including comparisons with Data Center Infrastructure Efficiency (DCiE), and delve into the specific nuances of measuring PUE in data centers. Additionally, Dgtl Infra discusses various strategies for improving PUE, highlighting everything from data center design to the impact of geographic location.

What is PUE (Power Usage Effectiveness)?

Power Usage Effectiveness (PUE) is a metric that measures the energy efficiency of a data center. To calculate PUE, divide the total amount of power entering the data center by the power used specifically by the IT equipment, such as servers, storage systems, and networking equipment. This metric helps quantify the proportion of a data center’s power utilized directly for computing versus the amount consumed by cooling systems, electrical losses, lighting, and other support infrastructure.

PUE Power Usage Effectiveness Representation for a Data Center shown by Windows Overlooking Green Hills and Sunset

By definition, PUE values are equal to or greater than 1. A PUE close to 1 indicates a high level of energy efficiency, signifying that the majority of the power is being utilized for computing. On the other hand, a PUE value of 2 suggests that for every watt of power used by IT equipment, an additional watt is needed for cooling and power distribution to the IT gear.

The Green Grid, a consortium of IT companies, data center operators, and professionals from various organizations, plays a crucial role in standardizing PUE measurement in data centers. Furthermore, these measurement protocols are incorporated into the ISO/IEC 30134-2 standard, which outlines guidelines and specifications for data center energy efficiency and performance metrics.

It’s worth noting that “Power Usage Effectiveness” is the widely accepted term, although it is occasionally referred to as “Power Utilization Effectiveness” within the data center industry.

Importance of PUE in Data Centers

Power Usage Effectiveness (PUE) is a critical metric for several reasons:

  1. Assessment of Energy Efficiency: PUE provides a straightforward metric for assessing the energy efficiency of a data center. It highlights how much additional energy is consumed for cooling, power supply losses, lighting, and support infrastructure. This insight helps data center operators understand how effectively energy is used
  2. Cost Reduction: By improving PUE, data centers can significantly reduce their energy consumption, leading to lower electricity bills. Given that energy costs are among the largest expenditures for data centers, enhancing efficiency has a direct positive effect on profitability
  3. Environmental Impact: Lowering the PUE of a data center reduces its carbon emissions, contributing to environmental sustainability. Essentially, a lower PUE indicates that a higher percentage of a data center’s energy consumption is being put to productive use, rather than being expended on cooling systems, power supply inefficiencies, and underutilized IT equipment
  4. Benchmarking and Performance Tracking: PUE allows data centers to benchmark their performance against industry standards and competitors. Tracking PUE over time helps in identifying trends, making informed decisions about investments in energy efficiency, and monitoring the impact of those investments
  5. Capacity Planning: An understanding of PUE facilitates more informed capacity planning. By knowing how much energy is spent on non-computing activities, data centers can plan upgrades or expansions more judiciously, avoiding unnecessary energy consumption increases
  6. Competitive Advantage: A low PUE can enhance a data center’s reputation, by demonstrating a commitment to energy efficiency and environmental stewardship. This can attract customers who prioritize sustainability and cost-effectiveness in their selection of data center services
Importance of PUE for Data Centers with a Backdrop of Glistening Sunlight that Filters Through Trees Illuminating a Globe

PUE interests a wide range of stakeholders, including data center operators, IT and network engineers, facility engineers, customers that rely on data center services, and regulatory agencies.

How is PUE Calculated?

Power Usage Effectiveness (PUE) is calculated by dividing the total amount of energy consumed by the data center (in kilowatt-hours, kWh) by the energy consumed solely by the IT equipment, such as servers, storage systems, and networking equipment, also in kWh. Since PUE is a ratio of two quantities measured in the same units (kWh), it is unitless. Typically, PUE is calculated annually to provide a consistent basis for comparison.

The mathematical formula to calculate PUE is as follows:

Here’s an overview of the components involved in the calculation:

Important Components Involved in the Calculation of PUE Power Usage Effectiveness Measurement Cooling IT Load
Source: The Green Grid.
  • Total Facility Energy: This is the sum of all the energy consumed by a data center, including:
    • IT Equipment: Energy used by all IT hardware, including servers, storage systems, switches, and routers, which are directly involved in computing and data storage tasks
    • Cooling Systems: Energy consumed by cooling systems to maintain optimal temperature and humidity levels for IT equipment. Components include chillers, chilled water pumps, fans, condenser water pumps, computer room air conditioning (CRAC) units, computer room air handler (CRAH) units, and humidifiers
    • Power Delivery Systems: Electrical losses that occur in the power distribution systems. This covers transformers, switchgear, uninterruptible power supply (UPS) systems, batteries, power distribution units (PDUs), backup generators, and electrical cabling
    • Lighting: Energy used to illuminate the facility
    • Support Infrastructure: Energy consumed by ancillary systems such as security, data center infrastructure management (DCIM) monitoring tools, network management systems, and any office space associated with the data center
  • IT Equipment Energy: This is the energy consumed solely by the IT equipment, excluding energy used for cooling, power delivery system electrical losses, lighting, and support infrastructure

Interpretation of PUE

  • Ideal PUE: The optimal PUE value is 1.0, indicating that all the energy consumed by a facility is dedicated solely to IT operations, without any waste on supporting infrastructure. However, achieving a PUE of 1.0 is practically impossible due to inevitable energy losses from power conversion and distribution inefficiencies. Consequently, operational data centers report a PUE greater than 1.0
  • Hyperscale Data Centers: These large-scale facilities generally report a PUE ranging from 1.05 to 1.2. Their energy efficiency is due to their immense scale, geographic location in cooler climates, and by operating at higher utilization rates
  • Modern Data Centers: Target a PUE between 1.2 and 1.5, reflecting improvements in efficiency compared to the industry average, which has historically exceeded 1.5. This industry data is supported by annual surveys conducted by the Uptime Institute
  • Transitional Data Centers: These facilities typically exhibit a PUE ranging from 1.5 to 2.0. This category includes facilities that have either not upgraded their infrastructure recently or have only partially implemented energy efficiency measures, and thus have not achieved the advanced efficiency levels characteristic of modern data centers
  • Inefficient Data Centers: These older sites exhibit PUE values as high as 2.0 or 2.5, indicating substantial inefficiency. Such high values require an audit to pinpoint the underlying causes of excessive energy use

Overall, a lower PUE signifies higher energy efficiency, showing that a greater portion of the consumed energy is being used directly for computing tasks, rather than going towards cooling systems, electrical losses, lighting, and other support infrastructure. Conversely, a higher PUE indicates lower energy efficiency, with more energy expended on non-computing functions.

PUE vs DCiE

PUE (Power Usage Effectiveness) and DCiE (Data Center Infrastructure Efficiency) are both metrics used to evaluate the energy efficiency of data centers. DCiE is the inverse of PUE and is expressed as a percentage. It quantifies the proportion of energy that is directly used by IT equipment out of the total energy consumed by a data center. The formula to calculate DCiE is as follows:

Alternatively, DCiE can be determined as the reciprocal of PUE:

A higher DCiE percentage reflects greater efficiency, indicating that a larger portion of the data center’s energy is used for IT operations rather than going towards cooling systems, electrical losses, lighting, and other support infrastructure. Ideally, DCiE should be as close to 100% as possible. Conversely, PUE is a ratio where lower values (approaching 1) signify higher efficiency.

While organizations generally prefer using PUE for its straightforward depiction of potential inefficiencies, DCiE is valued for its intuitive presentation of efficiency in a percentage format, making it easier for some to understand the effectiveness of energy use in data centers.

Example PUE and DCiE Calculation

Consider a data center that consumes a total of 100,000 kilowatt-hours (kWh) of electricity for all its operations. This total consumption includes everything from cooling systems and lighting to running the actual IT equipment such as servers, storage systems, and networking devices. Out of this total energy consumption, 80,000 kWh is specifically used to power the IT equipment, which is the core function of the data center.

Breakdown of Data Center Energy Use in Pie Chart Format for PUE and DCiE Calculation

Using this example, the formula to calculate PUE is as follows:

A PUE of 1.25 means that for every 1 kWh of electricity used by the IT equipment, an additional 0.25 kWh is used for other facility needs (like cooling and lighting). This indicates that the data center is using 25% more power than what is strictly necessary to run the IT equipment.

Applying the same numbers from the example, the formula to calculate DCiE is as follows:

This means that 80% of the total energy consumed by the data center is used for powering the IT equipment, while the remaining 20% is used for other infrastructure needs.

Measuring PUE in a Data Center

To accurately measure Power Usage Effectiveness (PUE) in a data center, it is essential to conduct extensive metering that captures the total power consumption of IT equipment as well as the supporting infrastructure, including cooling systems and lighting.

To accomplish this, it is necessary to install a significant number of power meters (ranging from tens to hundreds) throughout the facility. These meters are used to monitor the energy usage of IT equipment and the supporting infrastructure separately.

Electrical Engineer in Orange Safety Gear Inspects Control Panel which is a Power Meter

The detailed tracking of energy consumption is often supported by sensors that gather data, along with Data Center Infrastructure Management (DCIM) software. This setup enables accurate PUE calculations, offering a detailed overview and automated reports on the data center’s energy efficiency.

Key Components for Accurate PUE Measurement

To accurately measure PUE in data centers, it is essential to consider the following key components:

  1. Total Facility Energy: The initial step involves measuring the total energy consumption of the data center. This covers the power used by IT equipment, cooling systems, electrical losses, lighting, and other support infrastructure. Essentially, this measures all energy inputs into the data center facility, which may include electricity, as well as diesel or natural gas if these are used for generators or heating. The measurement is typically taken at the utility meter or via a dedicated meter specifically for the data center
  2. IT Equipment Energy: Subsequently, the focus shifts to measuring the energy consumption of IT equipment exclusively – covering servers, storage systems, and networking equipment. The energy usage of these components is generally assessed through meters at the Power Distribution Units (PDUs) that serve the IT equipment directly, but can also be measured at the Uninterruptible Power Supply (UPS) system or directly at the IT equipment input

Considerations for Effective PUE Measurement

To accurately measure PUE in data centers, the following specific parameters should be considered:

  • Units of Measurement: It is essential to consistently measure both the facility’s and the IT equipment’s power usage in terms of energy consumed to ensure both values are directly comparable. Typically, this is expressed in kilowatt-hours (kWh), rather than kilowatts (kW)
  • Measurement Period: The industry standard for calculating PUE is on an annual basis. This approach provides a holistic view that accounts for seasonal changes in cooling needs and varying IT load conditions throughout the year. While shorter measurement intervals, such as daily or monthly, can offer insights into trends, they must be interpreted with caution. This is because they can introduce seasonal biases, making the PUE appear more favorable or unfavorable than it actually is. For example, a data center’s PUE is generally higher in the summer due to the increased demand for cooling
  • Design PUE: The standard also encompasses a methodology for estimating the PUE of a facility that has not yet been constructed. This estimated or “design” PUE is calculated based on the anticipated energy consumption of the IT equipment and the efficiency of the planned data center infrastructure

Levels of PUE Measurement

The Green Grid has introduced a standardized framework for categorizing PUE measurements, known as “levels.” These levels are designed to classify PUE measurements based on their location, accuracy, frequency, and the methodology applied in gathering data on IT and facility energy consumption.

Levels of PUE Measurement from 1 2 3 Basic Intermediate Advanced in the format of a Chevron Process Chart

The categorization is based on the extent of metering involved and is defined as follows:

  • PUE Level 1 (Basic): Offers a simplified measurement approach, suitable for quick or initial assessments but with the lowest accuracy. IT equipment energy is measured at the Uninterruptible Power Supply (UPS) output and total facility energy is measured at the point of utility handoff to the data center
  • PUE Level 2 (Intermediate): Provides a balance between measurement accuracy and implementation complexity, using more precise data collection methods than Level 1. IT equipment energy is measured at the Power Distribution Unit (PDU) outputs, reducing the impact of UPS losses on the measurement. Similar to Level 1, total facility energy is measured at the point of utility handoff to the data center
  • PUE Level 3 (Advanced): Ensures the highest accuracy in PUE measurement, employing continuous, real-time data collection with high-quality metering equipment. IT equipment energy is measured at the point of connection to the IT equipment itself, providing the most accurate measurement of actual IT energy consumption. Total facility energy is measured the same as in Levels 1 and 2
MeasurementLevel 1 (Basic)Level 2 (Intermediate)Level 3 (Advanced)
Measurement AccuracyLeast preciseMore preciseHighest precision
IT Equipment EnergyUPS OutputsPDU OutputsIT Equipment Input
Total Facility EnergyUtility InputUtility InputUtility Input
Data Collection FrequencyMonthly/weeklyDaily/hourlyContinuous or very frequent intervals (15 minutes or less)

As shown in the table above, IT equipment energy is measured at various points from UPS output to direct IT equipment connection, while consistently measuring total facility energy at the utility handoff.

The diagram below also visually identifies the specific locations within an electrical distribution system where measurements for PUE calculations are taken. These locations pinpoint where the energy consumption of IT equipment and the total energy usage of the facility are assessed.

Electrical Distribution System Schematic Diagram Showing the Three Levels of PUE Measurement Labeled on Chart
Source: The Green Grid.

For a Level 1 (Basic) setup, IT equipment energy is measured at the UPS outputs, which is located higher up in the electrical distribution system compared to the PDUs or the IT equipment itself. This basic measurement provides a general overview of energy consumption on a monthly or weekly basis.

In contrast, Level 3 (Advanced) measures directly at the IT equipment input, offering a detailed account of the energy consumed by the IT hardware. This approach involves recording data at continuous or very frequent intervals – every 15 minutes or even seconds – ensuring a comprehensive and precise measurement of energy usage.

Because of this greater precision, the Level 3 (Advanced) PUE measurement involves higher costs and is designed for organizations that want to meticulously track every kilowatt-hour (kWh) of energy from its source to its usage.

However, by opting for Level 1 (Basic) measurements instead of Level 3 (Advanced), organizations might overlook significant cooling inefficiencies and electrical distribution losses. This could lead to the misleading perception that a facility is more energy-efficient under a Level 1 (Basic) assessment than it actually is when evaluated with Level 3 (Advanced) data.

Strategies for Improving PUE

Improving the Power Usage Effectiveness (PUE) of data centers can be achieved through several strategies: optimizing the design of the data center, selecting IT equipment that is highly energy-efficient, adopting advanced cooling technologies, adhering to rigorous operational practices, and selecting a geographic location that is favorable.

IT Professionals are Standing in a Data Center Server Room Holding a Laptop to Evaluate Different Strategies

1. Data Center Design

  • Hot Aisle/Cold Aisle Containment: This strategy involves separating hot and cold air streams to prevent the mixing of hot air expelled by servers with the cold air used for cooling. This separation significantly reduces the energy required for cooling, thus improving PUE
  • Blanking Panels: Installing blanking panels in unused rack spaces minimizes air leakage and directs airflow effectively towards active IT equipment. This method enhances cooling system efficiency, contributing to better PUE

2. IT Equipment Selection

  • Energy-Efficient Hardware: Choosing servers, storage systems, and networking equipment with high energy efficiency ratings, such as ENERGY STAR-certified products, reduces power consumption for IT loads, enhancing PUE
  • Virtualization and Consolidation: Consolidating workloads onto fewer, more efficient virtual machines (VMs) increases server utilization and decreases the need for physical servers and IT equipment. This reduced physical infrastructure approach lowers overall power consumption for computing and cooling, thus improving PUE

3. Cooling Technologies

  • Variable Speed Drives (VSDs): VSDs adjust the speed of fans and pumps based on the actual cooling demand rather than running them at full capacity constantly. This adjustment leads to lower electricity consumption for cooling and a more favorable PUE
  • Free Cooling Systems: Leveraging outside air for cooling, when temperature and humidity conditions are suitable, reduces energy consumption. Air-side economizers, for instance, use cooler outside air to provide “free cooling,” eliminating the need for mechanical refrigeration and thus improving PUE. The below chart illustrates that, with a constant IT load, utilizing free cooling can result in improved efficiency and consequently better PUE at lower temperatures
Chart Showing the Relationship Between PUE and Outdoor Ambient Air Temperature
Source: The Green Grid.

4. Operational Practices

  • Load Balancing: Optimizing workload distribution across computing resources ensures that no single server is overburdened while others are underutilized. This balanced utilization reduces energy consumption, thereby improving PUE
  • Raising Temperature Settings: Data center operators can slightly increase the temperature settings of cooling systems, such as CRAC or precision air conditioning units. Operating at higher temperatures reduces the energy directly used for cooling, thereby lowering overall energy consumption in relation to IT equipment usage and improving PUE

5. Geographic Location

  • Climate: Selecting locations with cooler climates minimizes cooling requirements and energy consumption by taking advantage of ambient air for free cooling. For example, the Pacific Northwest, including areas in Oregon and Washington, as well as Nordic Countries like Sweden, Finland, Norway, and Denmark, are key data center markets that offer a cool climate, making them ideal for natural cooling systems
  • Underwater Data Centers: These facilities utilize the natural cooling properties of ocean water, drastically reducing the reliance on traditional, energy-intensive air conditioning systems. The efficient cooling offered by underwater environments leads to improved PUE, compared to traditional land-based data centers

Comparing PUEs of Data Center Operators

The table below provides a snapshot of Power Usage Effectiveness (PUE) metrics from various data center operators who disclose this information. It is important to note that this data is self-reported and does not undergo a third-party audit process. Moreover, the figures are not standardized to account for definitional differences among companies. For example, some operators only consider their “stabilized” data centers when calculating their reported PUE metrics.

Data Center OperatorOperator TypePower Usage Effectiveness (PUE)
AlibabaCloud Service Provider1.22
CoreSiteRetail Colocation1.37
CyrusOneWholesale Colocation1.45
EquinixRetail Colocation1.46
GoogleCloud Service Provider1.10
Meta PlatformsInternet Company1.08
MicrosoftCloud Service Provider1.18
Switch, IncHybrid Colocation1.18
Data Center – Average1.58

As indicated in the table, retail and wholesale data center operators typically report higher PUE values compared to hyperscalers, such as cloud service providers (CSPs) and large internet companies.

Despite these differences, all data center companies share a common goal: to enhance their energy efficiency. Many have set specific objectives to lower their PUE, with targets for both the construction of new data centers and by upgrading existing facilities.

Mary Zhang covers Data Centers for Dgtl Infra, including Equinix (NASDAQ: EQIX), Digital Realty (NYSE: DLR), CyrusOne, CoreSite Realty, QTS Realty, Switch Inc, Iron Mountain (NYSE: IRM), Cyxtera (NASDAQ: CYXT), and many more. Within Data Centers, Mary focuses on the sub-sectors of hyperscale, enterprise / colocation, cloud service providers, and edge computing. Mary has over 5 years of experience in research and writing for Data Centers.

LEAVE A REPLY

Please enter your comment!
Please enter your name here