Sean Farney, Director of Data Center Marketing, Kohler Power Systems, highlights the need for an all-encompassing and standardized way to measure and report on data center sustainability across the entire data center ecosystem.
Sean Farney, Director of Data Center Marketing, Kohler Power Systems
In The Power of the Negawatt: Efficiency Improves Data Centers’ Energy Impact, Rich Miller recently gave the data center industry some long overdue credit for bending the power consumption curve down while significantly expanding server fleet size over the last 10 years.
Led from the front by the GAFAM Hyperscalers, we warmly embraced (Power Usage Effectiveness) as a measurement tool and, subsequently, gamified it amongst peer companies in an asymptotic race to 1.0.
While at Microsoft, at the behest of luminaries Mike Manos and Christian Belady (PUE’s primogenitor), I led a very talented team in bringing the first large-scale (120MW) Container data center online. By thinking creatively- putting servers in a box to more efficiently manage air flow- we achieved industry-leading PUE of 1.2. Not necessarily revolutionary; data center managers had been constructing all manner of contraptions to steer hot air in raised floor environments. But evolutionary to the extent that tetrapods transitioning from water to land was; kind of a big change. Not only did we drop OpEx 30%, but we advanced sustainability substantially by consuming less energy. From the Green Grid to the Wholesale building frenzy that ensued, this ethos spread across the entire industry, leading to a culture of PUE-guided sensibility.
As parallel in broader industry, I point to the success that Ford has had with their EcoBoost engine platform. By sizing the engine for median power output, but turbo charging it for brief periods of peak output- extremely evolutionary, but not revolutionary thinking- Ford has engineered 30% better fuel economy and 15% less greenhouse emissions. And as a customer, I’m delighted that my F-150’s “little” V6 puts out 450 horsepower yet gets 20 miles to the gallon, meaning fewer outings to Kwik Trip.
It’s useful to use this context when thinking about the next chapter in data center sustainability measurement. PUE hasn’t become passe, but thanks to increased measurement capabilities across the entire data center sustainability value chain, it now feels unidimensional. As an example, I recently saw an advert from NTT Data Centers that included a facility-specific breakdown of utility power generation source- gas, solar, wind, nuclear, coal, etc. This granularity is both impressive and a strong demand signal for multivariate measurement.
We really do have a great story to tell about our sustainability evolution in the space and shouldn’t waste the opportunity!
EPA Certified Tier 4 Final generator (Image: Courtesy of Kohler)
Similarly, diesel generator manufacturers have been engineering hard since 2014 to meet the U.S. EPA emissions requirements referred to as ‘Tier 4’ . Between aftermarket treatment technologies and enclosure systems, pollutant emissions have been drastically curtailed in the last ten years. The abatement systems employed also give operators the ability to measure and trumpet these advances. At Kohler, we are on the front-lines of this battle against emitted pollutants and are proud to provide factory-direct Tier 4 packaged units.
Given the increasing intensity of sustainability optics and the volumes of analytics we now have at our disposal, we really need a new measurement spec. I’ll playfully coin the moniker GUE- Green Usage Effectiveness (pausing for eyeroll) as simply a placeholder to express the need for an all-encompassing and standardized way to measure and report on the sustainability across the entire data center ecosystem. Answering the inevitable and immediate hue and cry, I say “we did this before”— think Uptime Institute Tier Classifications. It was a similar challenge, albeit with a design & engineering answer, but the industry agreed to common language and parameters to drive standardized outcomes. Just a thought, but one perhaps industry leadership should investigate. We really do have a great story to tell about our sustainability evolution in the space and shouldn’t waste the opportunity!
Sean Farney is the Director of Data Center Marketing at Kohler Power Systems.
Data Center Frontier is edited by Rich Miller, the data center industry’s most experienced journalist. For more than 20 years, Rich has profiled the key role played by data centers in the Internet revolution.
Rather than look to re-invent the wheel why not simply use the already published standardised KPI’s in the ISO 30134 series?
This includes the formal globally standardised definition of PUE as well as Renewable Energy Factor (REF), Energy Re-use Factor (ERF) as well as Water (WUE) and Carbon (CUE) Usage Effectiveness.
We already have what you are suggesting and importantly they are linked to international standards widely used by the industry such as ISO 9000 and ISO 27000 etc.
We justifiably have repeated calls for new metrics beyond PUE but sadly few are making use of what is already available.
Interesting article.. There are no silver bullets to suddenly become green or sustainable. Sustainability is on ongoing work in progress for both the data center and other industries..
I believe that the path for addressing utility outages by data centers will still require back-up generators for many more years. That being said, the hyperscalers have been can are continuing to leading the way, by testing energy storage as an alternative back-up system.
I believe it may be more cost effective to combine lower cost short term ride-though energy storage (30-60 mins) combined with tradition back-up generators
I think the big missing link is dealing with how water and electricity can both be used to achieve cooling, but affect different metrics (WUE and PUE). The race to make PUE look good (the gamification that Sean mentioned) led many operators to “burn” water instead of power for cooling because it’s invisible to PUE, even in water-stressed areas like Phoenix. The old school thinking was that burning water onsite probably was offset by the water consumed at the big thermoelectric coal/oil/gas plant upstream, but as the grid shifts to more water-efficient thermoelectric and water-free wind and solar, this assumption grows weaker each year.
We need some way of combining total cooling source into a single metric, regardless of whether it’s done with water- or air-cooled chilling. I guess one way would be to develop some embodied carbon conversions for water (MTCO2/kL) or embodied energy (kWh/L), but that undervalues water where it is scarce as opposed to where it is plentiful
Perhaps we just need to make a better game: depart the world of physics and just make a score that we want to lower. Perhaps a liter of water in Extremely High Water Stress areas (as described by the Aqueduct Water Risk Atlas) is worth 4 points in a score, while it’s only worth half a point in Low Water Stress areas. Anything to give an apples-to-apples comparison of data centers that choose between water or electricity while taking into account the local conditions.