In my last column I discussed the proposed ASHRAE 90.4 Data Center Energy Efficiency standard. At the time the article published, the second revision for comment (September 2015) mandated a PUE ranging from 1.3 to 1.6, dependent on its location in the 18 geographic zones. Needless to say, the data center industry was perturbed by the implications of a mandatory PUE as part of building standards. Since then, the third revision for public comment was released January 29, 2016. It completely removed all references to PUE, presumably in response to public criticism and comments from major colocation providers as well as hyper-scale cloud and internet heavyweights. The third revision has other complex energy efficiency compliance requirements in lieu of PUE, but we will have to see what happens next, since 90.4 may have further revisions or become finalized in the near future.
Although the 90.4 committee removed the PUE metric in March, the White House announced that under the new data center optimization initiative (DCOI), all federal data centers must reduce their PUE to under 1.5 by September 2018 (unless they are already scheduled to be shutdown as part of the Federal Data Center Consolidation Initiative [FDCCI]).
Nonetheless, for virtually any new data center being built, energy efficiency is no longer an afterthought, with or without PUE references in the 90.4 standard. PUE numbers have come way down for all new designs in general. Today, a PUE of 1.2 or less is the new normal for the hyper-scale data centers, and a PUE of under 1.5 for a newly constructed enterprise or colocation facility is no longer a dubious claim. Nonetheless, even as we approach the utopian PUE of 1.0 (zero facility energy loss), we have only just begun to see the other side of the mountain, as the industry proudly stands staring at the pinnacle of its accomplishment (or perhaps it is just the tip of the melting iceberg).
It is easy to overlook that even at a PUE of 1.0 every MW would still emit the same amount of heat into the environment as a load bank. So what can be done to address this proverbial Catch-22? The Green Grid introduced the energy reuse effectiveness (ERE) metric in 2010, as well as the related energy reuse factor (ERF). However, these reuse metrics are focused on recovery of waste heat energy to be used outside the data center whitespace or support spaces, thus reducing other energy demands, such as general building heating or hot water (or to reduce the non-data cooling load by driving an absorption chiller if feasible).
While data center waste heat energy reuse is an important step forward, it has been slow to take hold. It also has led to confusion regarding the PUE metric resulting in erroneous PUE claims of under 1.0 (e.g., a PUE of 0.9 if 10% of waste heat was reused). The energy reuse metric is defined as (ERF = reuse energy/total facility energy) and that ERF can only range from 0 to 1.0. The value of 0.0 means no energy is reused, while a value of 1.0 means all of the energy brought into the data center is reused. ERE is defined as [ERE = (1-ERF) × PUE], also note that as ERF goes to 0 (no energy reuse), ERE will equal PUE. In the example above, the ERF would be 1.1. Yet even if the ERF was 1.0 (all energy reused), it is important to understand that the PUE can never be lower than 1.0 regardless of how much energy is reused (which is clearly stated in the PUE metric).
There are a multiplicity of efforts by ASHRAE, the U.S. Green Building Council (USGBC, source of the LEED® standards), as well as the federal government and state agencies, to improve energy efficiency in commercial and residential buildings, with the ultimate goal of achieving “net zero” or “zero energy” building. This is where things begin to get “fuzzy.”
Before we get to the net zero energy data center (NZEDC), let’s see what’s happening outside our microcosm, as the net zero energy building (NZEB) or zero energy building (ZEB) requirement develops. California has already set a defined ZEB goal for new commercial buildings by 2030. ASHRAE has been peering through their own looking glass when they first released their “Vision 2020” report in 2008. They focused on many of the existing state of the art practices, using already prevailing technologies, stating, “Airtight insulation, high-efficiency elevators, designs that prefer natural lighting to artificial lighting, and solar water heaters on the roofs of many buildings, all optimize energy use and make these buildings more efficient.” However, while energy efficient building materials and methods clearly can substantially reduce energy use, it still does not get down to net zero.
In addition, the U.S. Department of Energy issued “A Common Definition for Zero Energy Buildings” in September, 2015 (http://1.usa.gov/1W6tac6).
The guiding principles of the U.S. DOE report is to “Create a standardized basis for identification of ZEBs for use by industry,” “Be clear and easy to understand by industry and policy makers,” and “Set a long-term goal and be durable for some time into the future.” It also defines the terms, “A zero energy building (ZEB) produces enough renewable energy to meet its own annual energy consumption requirements, thereby reducing the use of non-renewable energy in the building sector.”
The DOE report further states, “The team reached the conclusion that the word ‘net’ did not add substantive meaning to the name, since the definition fully describes how to account for delivered and exported energy. Therefore, in striving for simplicity, consistency and to accentuate the core objective, DOE and NIBS selected the term Zero Energy Building (ZEB). However, it is recognized that the terms Net Zero Energy (NZE) and Zero Net Energy (ZNE) are in wide use and convey the same meaning as Zero Energy. During the review process, the Project Team identified the need for additional definitions for related groupings of buildings. The team included definitions for ‘Zero Energy Campuses,’ ‘Zero Energy Communities’ and ‘Zero Energy Portfolios’ to expand the reach of the ZEB concept.”
While intended to make these issues and terms clearer, the report just begins to highlight how complex this will become for regular buildings, so how will this impact data centers if they are not excluded or separately recognized as a special category?
In the movie Back to the Future, Professor Emmett Brown said, “Roads? Where we’re going, we don’t need roads.” To paraphrase that thought, “Utility power? Where we’re going we don’t need any utility power.” And while Professor Brown was able to manipulate the space-time continuum by finding a way to generate 1.21 gigawatts to charge a “flux capacitor” by using a cold fusion power source (“Mr. Fusion,” which was the size of a coffee maker), ever larger hyperscale data centers may soon start using that much power in the not too distant future.
In the event that kitchen appliance size cold fusion gigawatt power generation does not become viable by 2030, what will constitute the NZEDC? I believe that one of the key elements to the zero energy data center will be based on substantially improving the waste heat energy recovery technology. While the majority of IT equipment today is air cooled, liquid cooled IT equipment is not just technically feasible, this was how the first mainframes were cooled and now liquid cooling has returned with many improvements. In fact, some models of modern mainframes and supercomputers, as well as a wide variety of servers, are available today that use liquid cooling. Moreover, The Green Grid has formed a Liquid Cooling Technology workgroup, which is expected to release a white paper later this year.
While in most cases, liquid cooling applications are currently being driven by the demand to increase CPU performance and package density while reducing cooling energy. In the long term, it may be the waste heat recovery and energy reuse that may lead to the ZEDC. Unlike air cooled IT equipment, liquid cooling can deliver higher grade heat, which allows for greater opportunity for energy recovery. While still in its early stages, some liquid cooled systems can be “cooled” by 120°F supply water and deliver even hotter return water (or other fluids), which allows for more effective heating applications. There is even the possibility of electrical power generation (via heat engine or other conversion technologies) in the not so distant future. This could then be used to power a portion of the IT load, which would get us much closer to the ZEDC.
Clearly the devil is in the details, as well as the semantics. I believe that “net” is the elephant in the room, since the difference between a ZEB and a NEZB is the ability to generate excess power back into the grid whenever the building requires less energy than it generates. In fact, in February 2015, ASHRAE responded to the request for comments before the DOE ZEB report was finalized, stating, “Do Not Remove the Word ‘Net’ from the Term ‘Zero Energy Building.’”
Net metering has been in use for many years for residential and commercial building. As its adoption increases, it may become a problem for the utility (and the grid). The existing grid was never meant to support a bidirectional, highly asymmetric power flow to-and-from endusers. Most typical commercial buildings use more energy during the day when they are occupied, and much less in the night. In those cases, solar generation can help offset that type of peak loads. In contrast, data centers draw a relatively continuous and significant amount of power on a 7x24 basis. If they were to use onsite solar panels to generate energy eight hours a day, they would then need to push twice as much power back into the grid over the next 16 hours to become a “net” zero site. A true zero energy building would be able to generate sustainable energy (not just generate energy locally) and also be able to store enough energy to minimize or eliminate the need to feed it back to the grid.
Energy storage will be another of the key technologies for the ZEDC. Many organizations are working on energy storage for grid, micro-grid, and building level applications. Even Tesla Motors has been developing high-capacity battery technology to increase solar residential applications, as well as larger scale data center applications, based on the latest battery technology used in their car.
In an even stranger view of the future of energy storage for the data center, Eaton and Nissan announced at the Paris COP21 Climate Summit, that they are going to offer a line of UPS systems based on used batteries from the first generation Nissan Leaf electric car. According to Nissan, each Leaf has a 24 kWh battery, which will reach the end of its “first life” after five years. In its “next life” as a UPS battery, it’s supposed to still have about 80% of its capacity so it can be used for “less demanding applications” (no, I am not making this up). Setting aside the repurposed batteries, the system is meant to directly accept DC power to the DC battery buss (from solar and wind) to have integrated sustainable energy generation for small scale data centers in areas with limited or poor utility power.
However, to date most of the other high-profile announcements about data centers that have gone solar (or other sustainable energy sources such as wind) have done so by means of the power purchase agreement (PPA). This is simply a contractual issue; the data center still receives all its power from the utility as it normally would. Although not a technical innovation, this is still a positive step, since it helps fund and increase the use of sustainable power generation for data centers, as well as many conventional loads by other organizations.
Even though energy efficiency edicts are well intentioned, the data center industry must be more actively aware of any pending energy efficiency recommendations specific to or inclusive of data centers from ASHRAE, USGBC, or other organizations, as well as government agencies. These can become mandates or standards, which could be enforced by local building departments, and could impede the technological innovation which could have improved efficiency. The data center and IT equipment have vastly improved their energy efficiency and performance over the past 15 years for self-serving interests; it lowers energy costs and allows greater computing without increasing power demands. It is now foreseeable that the zero energy data center (net or otherwise) may be closer to reality in the future, perhaps by 2030, even without Mr. Fusion.