In my last column, I pondered how the 50th anniversary of Earth Day and the second International Data Center Day might clash or align in their goals. However, since then, we have all become engulfed by the COVID-19 pandemic and subsequent lockdown. While stay-at-home orders have clearly highlighted just how much our society depends on our digital infrastructure, it’s important not to let the critically of the industry overshadow the effect it has on the environment. So, how do we reconcile the goals of Earth Day — mitigating climate change and pollution versus the general negative perception of massive data centers consuming terawatt hours of energy?
One of the tools that help improve data center efficiency is the creation of metrics and standards that define what we measure and how to calculate energy efficiency. Some data center equipment is relatively simple — for example, an electrical component, such as a transformer or a UPS with output/input expressed as a percentage. This can be calculated in a test lab based on power (kW) across the operating load range (0% to 100%) of the device rating. For cooling systems, it is far more complicated. While each device could be tested individually, there are multiple interrelated equipment factors, such as CRAC/CRAH, humidity controls, pumps, and chillers, not to mention the correlation of climate conditions on heat rejection system performance.
While all this is fine from a design and equipment purchasing perspective, it is only a rough indicator of actual operation conditions over the course of a period of time, such as a year. In a data center, operating conditions are constantly changing, and that’s an understatement. Moreover, power measurements (kW), even at regular intervals, are not as accurate or as meaningful as actual energy (kWh) consumption over one month or a year.
That brings us to the topic of efficiency metrics and standards. I recently had the opportunity to present a webinar for Mission Critical on the topic.
I would hope that if you are reading this, you have heard of The Green Grid (TGG) and its power usage effectiveness (PUE) metric, which was initially released in 2007. The PUE metric has been updated several times since then, and updated measurement methods have been clarified via a series of TGG whitepapers. Nonetheless, even today, I find that not all data center operators who monitor and report their PUE (publically or internally) are measuring it correctly. Moreover, it seems that many people are unaware that TGG transferred PUE and other metrics, such as carbon usage effectiveness (CUE) and water usage effectiveness (WUE) to the International Standards Organization (ISO) in 2018, and are now a part of the ISO/IEC 30134 series of key performance indicators (KPIs). While TGG and its mission continue to exist, it was absorbed into the Information Technology Industry Council (ITI) in 2019.
Over the years, many other organizations have put forth their own facilities-based efficiency metrics, but none have gotten as much traction as PUE. The bigger picture question is the IT equipment energy efficiency. This has been tried by many organizations, including TGG, but it has proven to be far more difficult to quantify. However, the U.S. Environmental Protection Agency (EPA) tried to address both the facility and the IT equipment with the Energy Star (ES) program for data centers. This was started in 2009 for the facility and was closely aligned with PUE but requires submitting a 12-month certified energy audit for the facility in order to be awarded the Energy Star certification for a given prior year. The basis of the award criteria is to be in the top 75th percentile of the qualified submissions, with the first award issued in 2010. Each subsequent year’s data had to be submitted again in order to be listed for that year. The program included data center equipment beginning with servers and then added storage systems in 2013 and large network equipment in 2016. In addition, the UPS certification was also added in 2016 and updated in 2019. The programs are all voluntary, and, to date, there are only 168 certified data centers listed.
Conversely, equipment vendors are far more active and very motivated to get their equipment certified for financial reasons, since, by law, all federal agencies and many states require purchasing Energy Star equipment if it exists, unless they can offer a valid reason to use a noncertified product. This has also become a commercial preference, since the cost for ES products now carry a small cost premium, and IT energy represents a substantial operating cost.
ASHRAE started including data centers in 2010 with the 90.1 Energy Standard for Buildings Except Low-Rise Residential Buildings. The data center industry criticized this as too mechanically prescriptive. In 2013, it was updated based on overall efficiency performance. Finally, ASHRAE released the 90.4 Energy Standard for Data Centers in 2016.
Unlike the TGG/ISO PUE metric, ASHRAE standards can become a legal requirement. The 90.4 standard states, “Compliance with this standard is voluntary until and unless a legal jurisdiction makes compliance mandatory through legislation.”
It should also be noted that, unlike PUE, 90.4 uses climate zones as a factor for cooling system energy efficiency. The 90.4 standard was updated in 2019, which tightened the mandatory compliance requirements for mechanical cooling and electrical efficiency.
The Bottom Line
Data center power and cooling efficiencies have improved significantly and made substantial progress since TGG first introduced PUE. PUEs have dropped down to 1.1 for some hyperscalers and range from 1.3 to 1.5 for many new colocation facilities. However, IT power demands have increased massively since 2007.
In the grand scheme of things, it begs the question: Are the energy and resources consumed by digital infrastructure offset by the reduction in energy for the activities its replacing? For example, it has led to the huge reduction of information dissemination by paper-based publications.
And speaking of which, as I write this, it appears that Mission Critical magazine (like many other periodicals) will no longer be produced in paper format and will continue digital only. While I recognize that this makes business sense as well as improves sustainability and climate change goals, on a personal note, I must admit to being old school and have to say I will miss it.
Theses offsets include end-to-end processes. Resources, like trees and water, are used to create paper. And energy is required to print and distribute newspapers, magazines, and snail mail as well as to recycle them.
It has also substantially changed shopping and other daily habits of millions of people, from traveling to stores to ordering online (even before the pandemic). This example is a double-edged sword, since “free shipping” has caused a massive uptick in online ordering and the consumption of energy and other shipping-related resources (cardboard, paper, etc.).
So does that negate the argument that data centers and global sustainably goals are mutually exclusive?
The COVID-19 pandemic is the horrific catalyst that has proven the value of data centers and communications systems that form a robust digital infrastructure (and the eco-structures they create and support). However, going forward, it has also impacted how people and various organizations will think about work, education, and many other aspects of daily life.
So while data center facility efficiency metrics and standards are important, they are only tools. It is ultimately up to the designers and operators to use them — not just to meet regulations but to continuously improve their overall efficiency and mitigate climate change. This is not only good sustainability stewardship, but it also lowers facility-related operating costs.
Even more importantly, facility metrics do not look at the energy used by the underlying core of digital infrastructure; it is the IT equipment used to meet the demand for services and applications that drive the underlying demand for more and larger data centers. Therefore, it’s the service providers and the end users that utilize these services and apps and drive the massive energy consumption in the data centers. For that, there are many pros and cons, but there is no metric.