The Metrics of Going Green with PUE

This blog has been written by Chris Wozniak.

It is second in a series about sustainability. To read the first blog, click here.  Last time, Michail Clutter, Andrew Fanara, and Dave Cox shared how the winds are changing around sustainability in the data center. Board of Directors, government entities, shareholders, and customers look at a data center’s sustainability as a reflection of the data center’s values. But something we have witnessed is that the numbers used to show sustainability vary from data center to data center. As a result, skewed metrics are used, analyzed, and compared. In this week’s blog, we take a more in-depth look at what Power Usage Effectiveness (PUE) is, how customers have miscalculated this metric, and what Casne has done to help them.

Calculating Power Usage Effectiveness

PUE looks at hours of operation and kilowatts (kW) used in the entire data center and divides it by the energy used by IT equipment.

Looks like a simple equation, right? But to get the numbers to start this equation, you will need to do some research and math. Your answer to that equation will be a general understanding of your PUE. How do you know where precisely your power consumption is happening, or where the challenge areas are? With a data center, teams are sub-metering at the power strip level. One of our customers wanted to know the amount of power used by every outlet. This information helped our customer make meaningful decisions on power allocation for a better PUE. You can’t get more specific than that. 

What Does PUE Mean for Your Operation?

The first mistake we see is that this simple metric represents different things to different people, but these differences are never captured and understood within the entire organization. Are we comparing PUE from facility to facility? How meaningful is that comparison if they reside in different geographies, have different tier ratings, contain dissimilar compute infrastructure, or have entirely different operational objectives? In most cases, comparing PUE between facilities, and certainly using PUE for bragging rights between entire organizations (we’re looking at you, Facebook and Google) is a fool’s errand. PUE can be meaningful and useful when used to benchmark a single facility in order to measure progress toward efficiency improvements or compare a cluster of facilities to expose design advantages or problem areas. Sometimes the original intent of measuring PUE becomes clouded over time as people come and go, or as different stakeholders get their hands on the numbers without the necessary context to understand their meaning.

Mistakes in Calculating PUE

When working with some of our customers, we have seen how they are incorrectly calculating this metric. Here’s a list of some of the ways we have witnessed inaccuracies occur:

• Human error: We have seen some of our customers make errors varying from wrong data used to inaccurate math. Additionally, when new assets go online, PUE calculations are often not updated, leaving your facility with an inaccurate PUE metric.
• Lack of instrumentation: In many cases, we have seen data centers lack the instrumentation needed to measure PUE effectively. If there are significant gaps in measurement, the accuracy of the PUE metric quickly loses value.
• Snapshot vs. continuous time-series data: Data fidelity is critically vital in measuring an accurate and meaningful PUE value. PUE that is calculated once per day when nobody is online or when the outside air is coldest (and free air cooling is most likely) essentially games the metric, and it again loses its usefulness. PUE should be measured with measurements taken at regular intervals throughout each day and every day of the year to provide a full profile of the facility’s performance over a wide timeframe.
• Outdated Power One-Lines: In the fast-paced construction process for data centers, it’s common for facility drawings to fall out of sync with changes to the electrical power infrastructure. As a result, the drawings used to identify the correct points of measurement can mislead the implementation and produce inaccuracies in the metric. This one is tough to catch without diving into minute details of the physical infrastructure, so we see many incorrect PUE values and misleading trends going up to the C-suite.
• Internal Pressure: With the ever-growing pressures to go green, companies will often set internal targets for PUE. We have seen teams work to artificially make their KPIs aligned with their organization’s arbitrarily aggressive target. As a result, the metric is gamed to meet the target with little or no actual operational efficiency progress.

Data Drives Decisions

While the mistakes above are common, they shouldn’t keep happening. We have worked with our customers to achieve the correct means of measurement and data collection to make PUE measurements accurate and verifiable. The path to these solutions includes:

• Consulting and Training for accurate calculations: Our team works to understand our client’s needs and goals through consultation. After implementing improvements within a system, we train our clients’ teams to use them effectively.
• One-Line Analysis and Drawing As-Builts: Casne’s Engineering team has helped clients by analyzing their one-lines to make sure they are up-to-date and that the facility’s drawings are accurate.
• Instrumentation Audit: We review our client’s physical infrastructure to identify and fill gaps in instrumentation needed to correctly calculate PUE or identify problems with data quality in existing DCIMs, among other things. This process often exposes errors made and not caught during construction and errors in reporting due to under/over-reporting power usage.
• Integrate existing data: Our team works to take data already available in existing equipment, ensure its accuracy, and then integrate it into the facility’s existing DCIM or reporting systems. Many times, the instrumentation already exists to produce highly reliable PUE metrics but has not been integrated so that the data is available for reporting. Solving this can bring the considerable benefit of moving from unreliable and inaccurate manual readings to automatic and verifiable real-time calculations.
• Created a system to calculate and report PUE: Ultimately, calculating and reporting PUE-allows our clients to make more informed operational decisions. We can create facility performance profiles at different times of the year to inform operational decisions. Additionally, we can develop predictive analysis to shift workloads based on expected efficiency models.
• Sharing data via displays and dashboards: Casne has worked with clients to integrate this type of data onto their dashboards and heads-up displays. The dashboards complement the “single pane of glass” visibility of data center operations that we’ve helped some of our clients achieve on their data infrastructure platform. Clients can see real-time values of calculated KPIs, like PUE, along-side real-time statuses of data center facilities and assets.
• Asset Automation: It’s common for new assets to be frequently added to and removed from the data center. For PUE to be accurate, these assets need to be added to the data infrastructure platform as quickly as possible. We automate the addition of new assets into your platform to have the dual benefit of PUE accuracy and reducing the extensive hours spent on manual integrations.

When it comes to finding an accurate PUE to drive decision making, Casne (link) has 40 years of experience in professional engineering and technology integration for critical facilities. Our engineers and technologists develop and support engineered solutions using the best available products and technologies. Contact me here to discuss your data center’s energy management needs.