As the data center industry continues to expand and evolve, the metrics used by managers to optimize various operational aspects are also advancing. Among these metrics, Power Usage Effectiveness (PUE) remains the most popular. Widely adopted across the industry, PUE played a significant role in improving data center efficiency between 2007 and 2015, during which the average PUE decreased from approximately 2.5 to 1.6.

However, PUE is far from perfect. It has detractors as it is not an all-encompassing metric. Let’s examine the pros and cons of PUE, as well as other metrics that data centers should consider – not as a replacement for PUE but as a complement.

Power Usage Effectiveness

PUE is the ratio of total facility energy usage to IT equipment energy usage within a data center. The ideal, though unattainable, PUE is 1.0, where all power entering the data center is used solely for IT equipment with no waste. While PUE is useful for tracking the performance of a specific data center over time, it has limited value for comparing different data centers due to factors such as variations in calculation methods, unique configurations, climatic differences, and grid conditions, among others.

IT Metrics

Now let’s take a look at a few metrics that have emerged on the IT side:

• Rack density is a measure of how many kilowatts of power are available within a rack. Rack density averages are currently in 5 kW to 8 kW range but some data centers are heading to 100 kW and beyond.
• ARPA-E’s COOLERCHIPS program has tweaked PUE into what it calls Total Usage Effectiveness (TUE). It is the same as PUE except fan energy is subtracted from the IT equipment in the denominator.
• Performance Per Watt (PPW) measures the energy efficiency of every device in the data center. It is a really complex to calculate. It requires the assignment of a Performance Indicator (PI) to each device such as a testing benchmark and then figuring out the average device utilization rate against its wattage. This one is not for the faint of heart.
• Rack usage percentage time is the amount of time a rack is being used at full capacity. This one makes sense for data centers that plan to run close to full utilization much of the time, such as those investing heavily in AI that want to see a fast return on investment.

Sustainability Metrics

A series of metrics have been proposed to help data centers measure such things as their carbon footprint, the achievement of net-zero targets, and their overall level of sustainability. These include:

• Carbon Usage Effectiveness (CUE) measures total CO₂ emissions from data center energy against the overall energy consumption of IT equipment.
• Water usage effectiveness (WUE) addresses the fact that water is scarce in many parts of the world due to droughts or urban growth or because of regulatory pressure to be a more responsible environmental citizen. It measures the amount of water used to cool IT assets.
• Scope 1, 2, and 3 emissions are increasingly on the radar of data centers due to US Environmental Protection Agency (EPA) pressure. Scope 1 emissions are greenhouse gases (GHGs) released directly from fuel combustion in boilers, furnaces, diesel generators, cooling equipment, or vehicles. Scope 2 emissions are indirect GHGs due to purchased energy such as that used for electricity, steam, heat, or cooling. Scope 3 emissions extend responsibility for emissions across the organization’s supply chain.
• Renewable Energy Factor (REF) is a calculation of the proportion of energy consumption coming from renewable sources.
• Energy Reuse Effectiveness (ERE) measures effectiveness in redirecting energy to other uses in the facility, such as using a heat exchanger to convert server discharge heat into heating for surrounding buildings.
• Work Per Unit of Energy concerns the average capacity utilization of server equipment and is a way to show that energy use is productive and is not wasteful in terms of the data center’s carbon footprint.

Facilities Metrics

There are a few good metrics directly related to the facilities side as well:

• Mechanical Load Component (MLC) is an attempt to consolidate all cooling, fan, pump, and heat rejection power and compare it to the total power fed into IT equipment represented as a ratio.
• Electrical Loss Component (ELC) quantifies losses across the electrical distribution system by comparing what comes in from the utility services to what is received at the IT cabinet.

The Metrics of the Future

No metric is perfect. Each has its pros and cons, and some are more relevant to certain types of data centers than others. The needs of highly dense, AI-driven data centers differ significantly from those serving organizations with modest IT requirements. Each facility should focus on the metrics that align best with its goals—whether productivity, efficiency, cost-effectiveness, sustainability, or a combination of these.

Metrics are an essential part of effective management, as they replace subjective opinions with objective data to assess progress and identify areas for improvement in the data center. While no metric is perfect, ignoring them can be a costly mistake. Identify the metrics that align with your operational needs and leverage them to achieve both your short-term and long-term goals.